A Professional Journal Published by The International Federation of Technical Analysts

12

Inside this Issue

6 Applying Head-and-Shoulders
Pattern on MACD Histogram Indicator
to Forecast Market Direction by
Mohamed A. Elaasar, MFTA
17 Trading Strategies Based on
New Fluctuation Tests by
Daniel Ziggel and Dominik Wied
21 An Introduction to W. D. Gann’s Time
Cycles by Mikael Bondesson

…the true speculator is one

who observes the future and
acts before it occurs. Like
a surgeon he must be able
to search through a mass of
complex and contradictory
details to the significant facts. 
Then still like the surgeon, he
must be able to operate coldly,
clearly, and skillfully on the
basis of the facts before him.

—Bernard Baruch
 The Future of
Technical Analysis
has arrived.

Discover a whole new world of possibilities at www.mav7.com/ifta

 IFTA JOURNAL 2012 EDITION

Letter From the Editor

by Rolf Wetzer......................................................................................................................................5

Papers
EDITORIAL
Regina Meani (ATAA, STA, APTA) Trading Strategies Based on New Fluctuation Tests
Editor
[email protected] by Daniel Ziggel and Dominik Wied................................................................................................. 17

Elaine Knuth (SAMT)

An Introduction to W. D. Gann’s Time Cycles
Editor by Mikael Bondesson.........................................................................................................................21
[email protected]
Moving Averages 3.0
Dr. Rolf Wetzer (SAMT) by Manfred G. Dürschner..................................................................................................................27
Editor
[email protected] Trading the T-Grid
by John Gajewski................................................................................................................................41
Send your queries about advertising
information and rates to [email protected] The Volatility-Based Envelopes (VBE): a Dynamic Adaptation to Fixed Width
Moving Average Envelopes
by Mohamed Elsaiid, MFTA...............................................................................................................57

MFTA Research
Applying Head-and-Shoulders Pattern on MACD Histogram Indicator to Forecast Market
Direction
by Mohamed A. Elaasar, MFTA.......................................................................................................... 6
Applying Trading Strategies to Price Channel Breakout
Trading—Statistical Significance of Channel Breakout Variation
by Robin Boldt, MFTA........................................................................................................................ 33
Is Average True Range a Superior Volatility Measure?
by Glenn Marci, CFTe, MFTA............................................................................................................. 45

Book Reviews
Display stock market charts in a street Alpha Trading: Profitable Strategies That Remove Directional Risk by Perry Kaufman
Hongkong, China—photo by Nikada Reviewed by Larry Lovrencic............................................................................................................ 63
The Evolution of Technical Analysis by Andrew W. Lo and Jasmina Hasanhodzic
Reviewed by Regina Meani.............................................................................................................. 64
Author Profiles................................................................................................................................. 65
Directors and Board........................................................................................................................ 66

IFTA Journal is published yearly by The International Association of Technical Analysts. 9707 Key West Avenue, Suite 100,
Rockville, MD 20850 USA. © 2011 The International Federation of Technical Analysts. All rights reserved. No part of this
publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying for
public or private use, or by any information storage or retrieval system, without prior permission of the publisher.

IFTA.ORG PAGE 3
IFTA2012 25th Annual Conference

Singapore Check the website for

updates: www.ifta.org
 IFTA JOURNAL 2012 EDITION

Letter From the Editor

by Rolf Wetzer

Dear IFTA Colleagues and Friends:

In the very beginning, Technical Analysts were called “chartists.” They were those
who plotted and analysed market price data in order to predict the markets. In the
general public perception, this was nothing other than reading tea leaves. Yet it was
in this period that many ideas were developed on how to graphically describe price
movements, and how to come to conclusions from these graphs. Since then, there has
been much development and change, and Technical Analysts borrow from disciplines
such as psychology, statistics, computer science, information theory or even physics. It is a forum where
Now, in the age of seeming unlimited computer capacity, every tool that helps analysing
market data qualifies as technical analysis. market technicians
The IFTA Journal depends entirely on the papers and articles that we receive from
our IFTA colleagues. To start, we do not have a plan or a theme that we want the journal from around
to follow, but instead allow the flow of submissions to determine the theme. For this
year, it turns out that the main theme is volatility. We have included different papers the world come
on that subject, from the very basic volatility breakouts to more complex treatments of
volatility and correlation relationships. Nevertheless, a very broad range of technical together and speak
topics are covered, including Gann analysis, heuristic pattern search within the MACD
indicator, and a specific method to define moving averages based on modern cycle the same language
theory.
The journal mirrors the true spirit of IFTA. It is a forum where market technicians
from around the world come together and speak the same language. You will read
papers from Australia, Africa, Europe and America. Some of the papers are abridged
versions of prize winning papers from local technical societies. To this, a special
thank you goes to the German society VTAD which helped to motivate their local prize
winners to rewrite their papers for us. The other papers are colleagues’ MFTA papers
with two book reviews from our Australian colleagues.
A number of changes occurred within the IFTA Journal this year. Regina Meani
(APTA, ATAA, STA), our editor for the last four years, stepped aside. Regina
reanimated the journal and it is due to her efforts and energy, we have this
internationally published platform where technicians can present their ideas. Thank
you very much, Regina.
This year’s journal was produced by a new team. I want to thank Elaine Knuth
(SAMT, AAPTA) for copy editing the papers in this journal. Without her help, dedication,
technical knowledge and hard work, this journal would definitely not exist.
I would also like to thank Linda Bernetich and Jon Benjamin for their input in
publishing the journal and last but not least, Michael Samerski (ATAA, APTA) and Mark
Brownlow (ATAA, APTA) for their part in the reading and selecting of the papers.
A lot has happened during the course of the year. Much news turned out to be just
volatility, but some rather severe events occurred in Japan this spring. Therefore, I
would like to dedicate this journal to our friends of the Nippon Technical Analysts
Association (NTAA). Let us remember that the very first Chartists came from Japan
where Technical Analysis was born.

IFTA.ORG PAGE 5
 IFTA JOURNAL 2012 EDITION

Applying Head-and-Shoulders Pattern on MACD

Histogram Indicator to Forecast Market Direction
by Mohamed A. Elaasar, MFTA

Introduction In this paper I will, first, define the Head-and-Shoulders

Moving Average Convergence/Divergence (MACD), pattern on MACD Histogram as well as the different types and
constructed by Gerald Appel is a well known and established forms of the pattern, which can be recognized on the chart, in
indicator in technical analysis. There is the MACD indicator and addition to showing how to recognize the pattern, while it is still
the MACD Histogram indicator. The MACD indicator consists forming on the chart.
of three exponential moving averages, two of them represent Secondly, I will illustrate the trading techniques to trade this
the MACD line, which responds quickly to price changes, while pattern and how to benefit from its appearance to detect the
the third line represents the Signal line, which responds slowly market direction and generate profits.
to price changes, and is made of the MACD smoothed with Finally, I will conduct a statistical analysis test for the Head-
another Exponential Moving Average (EMA). In this paper we and-Shoulders pattern on the MACD Histogram indicator to
are concerned with the Histogram form of the MACD indicator, examine the frequency of its appearance, as well as its degree of
rather than the normal MACD indicator. predictability.
John J. Murphy stated the following about the MACD
Histogram: “The Histogram consists of vertical bars that show Methodology
the difference between the two MACD lines.” 1 By subtracting This study uses daily data via Reuters© from diversified
the Signal line from the MACD line, the resulting difference is markets and equities to make statistical analysis on the Head-
then plotted in the form of vertical bars above or below a line and‑Shoulders MACD Histogram pattern. These markets are:
called Zero Line (Figure 1).
ƒƒ Dow Jones industrial average (.DJIA)
Figure 1: MACD Histogram indicator ƒƒ General Electric stock (GE)
ƒƒ Silver (XAG=)
ƒƒ British Pound Sterling (GBP=X)

These four instruments cover three major liquid markets:

U.S. equities (Dow Jones Industrial Average as a market index
and General Electric as an individual stock), commodities
(silver) and currencies (British Pound Sterling). These
instruments have long history and adequate examinable
amount of daily data.
To standardize analysis, daily charts and daily data for each
one of the four markets for the period starting January 1983
till December 2009 were used. This period of 27 years is the
only period that has reliable available daily data for all the four
examined markets.
Before 1983, reliable data for silver, British Pound Sterling
and General Electric are missing. Dow Jones data goes back to
Typically, all market technicians consider that when price is 1907 and is used used in one example.
trading below the zero line, it means we are trading in a down Various markets are examined to show that this pattern is
trend, while trading above the zero line is a sign of an uptrend. applicable in non correlated markets. Additionally, we measured
Building on this understanding of the MACD Histogram, and the differences in frequency of appearance of the pattern and
from my long experience as a technical analyst in markets, I will differences in profitability.
concentrate on one specific pattern, identified over the years This study examines the following:
during my work and observation. The pattern is a Head and
Shoulders on the MACD Histogram indicator and has proven to be ƒƒ Number of appearances of the Head-and-Shoulders pattern
a signal for market direction. on the MACD Histogram indicator during the period under
This paper will concentrate on recognizing and trading this examination;
pattern and verification of the Head-and-Shoulders pattern on ƒƒ Success/ Failure rate of the pattern;
MACD Histogram with a statistical analysis. ƒƒ Average Profit/Loss of the pattern;
ƒƒ Average time duration of the move.

PAGE 6 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

Types of Head-and-Shoulders patterns This pattern appears on the negative side of the MACD
on the MACD Histogram indicator Histogram indicator, which is below the Zero line (Figure 3). The
Inverted Head-and-Shoulders pattern signals an expected rise
The Head-and-Shoulders pattern emphasized here is not in the market.
the classic and traditional Head-and-Shoulders pattern that
appears on the price side of the chart, formed from the price Figure 3: Inverted Head-and-Shoulders pattern on MACD
bars or candlestick. Instead the pattern we examine appears Histogram indicator
on the indicator’s side of the chart and specifically on the MACD
Histogram indicator.
The MACD Histogram Head-and-Shoulders pattern is very
similar to the price Head-and-Shoulders pattern in appearance,
and its definition.
There are two forms of the Head-and-Shoulders pattern,
as follows:

A—Normal Head-and-Shoulders:
In this case the pattern appears on the positive side of the
MACD Histogram indicator above the Zero line.
We can use Dr. Alexander Elder’s definition of the Price
Bars Head-and-Shoulders that states: “The head is a price
peak surrounded by two lower peaks, or shoulders.” The
same definition can be applied to the pattern on the MACD
Histogram indicator, but with substituting the word
‘price’ in the definition with the words ‘MACD Histogram’s
Vertical Bars’, so it will read as follows: “The head is a ‘MACD The Head-and-Shoulders pattern that forms on the MACD
Histogram’s Vertical Bars’ peak surrounded by two lower Histogram indicator can appear in two types:
peaks, or shoulders”2 (Figure 2). Normal Head-and-Shoulders
pattern is a bearish sign for the market that means the market 1—Classic Head-and-Shoulders
will experience a price decline. In this type of the pattern, we can see the very classic form,
where the head falls between the shoulders (Figure  2).
Figure 2: Normal/Classic Head-and-Shoulders pattern on
MACD Histogram indicator 2—Separated Head-and-Shoulders
In this pattern, the head is separated from the shoulders.
The head maybe separated from one of the two shoulders,
while attached to the other shoulder (Figures 4 and 5). In other
cases, we can find the head separated from both shoulders
(Figure 6).

Figure 4: Separated Head-and-Shoulders pattern on

MACD Histogram indicator

B—Inverted Head-and-Shoulders:
As in the price pattern, the Head-and-Shoulders pattern on
MACD Histogram indicator can appear in the Normal Head-and-
Shoulders form, as well as the Inverted Head-and-Shoulders form,
defined as: The head is a ‘MACD Histogram’s Vertical Bars’ bottom
surrounded by two higher bottoms, or shoulders.

IFTA.ORG PAGE 7
 IFTA JOURNAL 2012 EDITION

Figure 5: Separated Head-and-Shoulders pattern on MACD Examples:

Histogram indicator
1—Developed Markets:
An example from developed markets is the British index
(FTSE 100 Index, Fig. 7) illustrating how the Inverted Head-and-
Shoulders pattern on the MACD Histogram lead a continuation
pattern of a major move in mid 2009.

Figure 7: FTSE 100 Index (.FTSE) – Daily

Figure 6: Separated Head-and-Shoulders pattern on MACD

Histogram indicator

Another example from developed markets is the French

index (LYXOR CAC40 Index, Fig. 8), which shows two examples
of a similar Inverted Head-and-Shoulders pattern with both
shoulders connected to the head at the middle. Each of the two
patterns foreshadowed an up move by the index by about 3% in
seven to eight days—the first pattern appearing in mid 2008 and
the second in October 2008.

Figure 8: LYXOR CAC40 Index (.INCAC) – Daily

From my experience, the Head-and-Shoulders pattern

appears on the MACD Histogram indicator in most markets,
independently of national exchange, and in all time frames.
The following are examples display developed markets like
USA, UK, and France, as well as equities from several emerging
markets, such as Egypt, Israel, Saudi Arabia, South Africa and
Brazil. Some equities charts are for individual stocks and others
are for market indices.
Additional examples will be provided from markets other
than equities markets, such as commodities and currencies.
The pattern can appear on any time frame from the minute
to hourly, to the daily, weekly and monthly charts. (The monthly
period, however, requires a relatively long period for the pattern
to develop and confirm.) We find this pattern appears on very
old charts dating back to the 1900s when stock markets were 2—Emerging Markets:
still establishing themselves in North America (an example The appearance of the Inverted Head-and-Shoulders pattern
will be shown later for the Dow Jones in year 1907) and even on the Egyptian stock (El-Ezz Steel Company, fig. 9) signaled the
in a more recent example such as the example of the Egyptian beginning of a retracement that lasted 11 consecutive sessions
Stocks Index (EGX30). (11 vertical MACD Histogram bars) from the date of entering
this trade in October 2008, generating around 15% profit at the
exit point.

PAGE 8 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

Figure 9: El-Ezz Steel Company (ESRS.CA) – Daily For the emerging markets in Middle East and Asia, we review
a daily chart for the Saudi Index (Tadawul, Faig. 11) as a example.
Figure 11 shows how the Inverted Head-and-Shoulders pattern
lead the upward retracement in February 2009 and lasted for 10
consecutive trading sessions and resulted in 13% gain.
Israel Chemicals (Fig. 12) shows how the pattern repeated
in the same cycle, resulting in terminating the original trend
direction and shifting to the alternate direction, each time the
Head and Shoulders of theMACD leading the directional move.

Figure 12: Israel Chemicals (ICL.TA) – Daily

Our next example from the emerging stock market of Egypt

is Orascom Telecom Holding (Fig. 10) , which experienced
a sharp increase in price by 22% in 9 days following the
formation of an Inverted Head-and-Shoulders pattern on
MACD Histogram indicator in mid 2009.

Figure 10: Orascom Telecom Holding (ORTE.CA) – Daily

Figure 13: Harmony Gold Mng (HARJ.J) – Daily

Figure 11: Tadawul All Share Index (.TASI) – Daily

Harmony Gold Mining listed on the Johannesburg stock

exchange (Fig. 14) is an example from smaller or emerging
markets of Africa. Here, we see the stock declined sharply after
the appearance of the Normal Head-and-Shoulders on the MACD
Histogram indicator.
The Sao Paulo SE Bovespa Index in Brazil (Fig. 14) moves
upwards upon the appearance or trigger of an Inverted Head-
and-Shoulders pattern in the beginning of 2004.

IFTA.ORG PAGE 9
 IFTA JOURNAL 2012 EDITION

Figure 14: Sao Paulo SE Bovespa Index (.BVSP) – Daily 4—Commodities

In Figure 16 we see an example of the Head-and-Shoulders
pattern on MACD Histogram indicator in commodities market.
As we see on the chart, here too, the pattern signaled the onset
of a significant decline for gold, starting in July 2008.

5—Time Frames
Daily charts were shown in the previous examples. The
following are weekly charts from developed and emerging
countries with the most recent example of the EGX30 at the end
of 2009.
Figure 17 presents a weekly time-frame in a developed
market. The Inverted Head-and-Shoulders sharply reversed
the trend after its formation in October 2002. The stock’s price
increased by 29% in 10 weeks.

3—Currencies Figure 17: Citigroup, Inc. (C) – Weekly

An example from the foreign exchange market is the
currency pair of Australian Dollar vs. U.S. Dollar (Fig. 15).
This chart shows an Inverted Head-and-Shoulders with both
shoulders separated from the head before the increase of the
Australian Dollar over the USD by the end of 2007.

Figure 15: AUD/USD (AUD=X) – Daily

Figure 18: Africa-Israel Investments Ltd. (AFIL01.TA) –

Weekly

Figure 16: Gold (XAU=) – Daily

An example in emerging markets from Israel. The formation

of normal Head-and-Shoulders with the right shoulder
separated from the head in March 2002 resulted in a decline in
the stock’s price.
The weekly chart (Fig. 19) of last few weeks of 2009 shows
how the Egyptian EGX30 index dropped following the formation
of a Classic Head-and-Shoulders on the MACD Histogram
indicator.

PAGE 10 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

Figure 19: EGX30 Index (.EGX30) – Weekly Trading the Pattern

First of all, we have to setup the MACD Histogram indicator
that we will use on the chart. The default settings of the MACD
Histogram indicator should be the same as the default set in
Reuters charts, used as the source of data and charting.

MACD Histogram Settings:

ƒƒ Short Periods: 12
ƒƒ Long Periods: 26
ƒƒ Signal Line Periods: 9
ƒƒ Averaging Method: Exponential

In other charting systems, the settings may appear in

different names, such as:

Figure 20: Dow Jones Industrial Average (.DJI) – Daily ƒƒ Fast EMA (Exponential Moving Average): 12
ƒƒ Slow EMA (Exponential Moving Average): 26
ƒƒ MACD or Signal EMA (Exponential Moving Average): 9

Figure 21: Forming Normal Head-and-Shoulders pattern on

the MACD Histogram indicator

6—Very Old Example:

This example could be considered one of the oldest evidences
on the appearance of the Head-and-Shoulders patterns on the
MACD Histogram indicator on technical charts, as the inverted
formation of the pattern appeared on the Dow Jones daily Chart.
The right shoulder was formed by the end of March 1907 and
pushed the index to increase 9.5% from 77.40 to 84.80 in 14 days.
Figure 22: Forming Inverted Head-and-Shoulders pattern
Using Head-and-Shoulders patterns on on the MACD Histogram indicator
MACD Histogram indicator
The Head-and-Shoulders pattern on MACD Histogram
indicator helps the technical analysts determine potential
market direction. The Normal Head- and-Shoulder pattern
suggests that the following few days will be a declining market,
while the Inverted Head-and-Shoulders form of the pattern
suggests a rise in the market. Using the pattern to determine
likely market direction in the near future is a simple, yet
efficient application of this pattern. Another application is to
use the described pattern as a confirmation indicator along with
other technical analysis tools.
Additionally, this Head-and-Shoulders pattern on MACD
Histogram indicator could be successfully traded.
We now present the trading techniques of this pattern
followed by a detailed statistical analysis showing the accuracy
of the Head-and-Shoulders pattern on MACD Histogram
indicator.

IFTA.ORG PAGE 11
 IFTA JOURNAL 2012 EDITION

Trading the Head-and-Shoulders pattern on MACD Figure 24: Trading example for an Inverted Head-and-
Histogram indicator starts by recognizing the pattern on the Shoulders pattern on the MACD Histogram indicator
chart, as it is forming, which is never as easy as recognizing the
pattern post its complete formulation. Following is a described
method for trade entry during pattern formation:
Figure 21 shows how the pattern will look on the MACD
Histogram vertical bars, while it is still forming in the case of
a Normal Head-and-Shoulders pattern, as well as showing the
entry bar to a trade.
Figure 22 shows the pattern on the MACD Histogram vertical
bars, as it is forming in the case of an Inverted Head-and-
Shoulders pattern. It also specifies the entry of the trade.

Entering the trade

After recognizing the Normal or Inverted pattern, one
can see that entering the trade will always be from the right
shoulder after spotting its peak. We will enter the trade during
the formation of the right shoulder and before the pattern is
completed. In the case of Normal Head-and-Shoulders pattern
(Figure 21), the right shoulder’s peak is the highest vertical bar Exiting the trade
of the MACD Histogram indicator in the right shoulder, which is Exiting the trade does not differ from the entry
followed by another bar to its right that closes lower. This latter methodology. Exiting the trade should be decided when a
bar is considered a confirmation of the formulation of the right reversing bar is formed.
shoulder. This is the signal to enter the trade at the opening price In the case of Normal Head-and-Shoulders pattern, where
of the following right bar. In short, entering the trade is made the slope of the bars is down, the signal to exit the trade will be a
at the opening price of the second lower bar to the right of the bar that reverses the slope to be an up slope. Once this happens,
highest bar of the right shoulder. A short (Sell) position will be the exit should be made at the opening of the second bar to the
made, as the market is expected to decline (Figure 23). right of the reversing bar (Figure 23).
In the case of Inverted Head-and-Shoulders pattern with an
Figure 23: Trading example for a Normal Head-and- up slope, exiting the trade should be decided at the opening price
Shoulders pattern on the MACD Histogram indicator of the second bar to the right of the reversing bar (Figure 24).
In Figure 23 we see a real case for the British Pound Sterling
(GBP) by the end of 1996, where we can recognize a Normal
Head-and-Shoulders pattern with the entry and exit points for a
short (Sell) trade spotted on the chart. A drop in the price can be
easily seen, which resulted in an approximate move of 3% of the
entry price in 10 days.
In Figure 24 between September-October 1985 the chart
shows General Electric stock forming an Inverted Head-and-
Shoulders pattern on the MACD Histogram indicator. The entry
of a long (Buy) position is shown in green. This position resulted
in a return of approximate 3.7% in eight days as price increased
following the formation of the pattern. The red line shows the
bar at which the trade was closed.

Stop Loss
When should a trade be considered a failure? In other
words, when should one exit the trade without making any
A similar method of entering the trade applies on the Inverted profit or with minimal loss? Depending on the aforementioned
Head-and-Shoulders patterns (Figure 22). We will locate the understanding, one should expect a signal based on the
deepest (lowest) vertical bar on the right shoulder that is movements of the vertical bars of the MACD Histogram
followed by a higher low bar, which is considered the entry indicator.
signal informing us that the following bar is the entry bar. The After entering a trade at the right shoulder and one of
entry is made at the opening price of the second higher low bar the following bars to the right of the entry bar moves in the
that follows the lowest bar of the right shoulder. In this case a reverse directions of the entry bar and the slope, we must
long (Buy) position will be made, as the price is prepared to rise first observe if the bars crossed the zero line to allow us to act
(Figure 24). according to the following cases:

PAGE 12 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

1. If the bars cross the zero line, the exit should be made at the Figure 27: Exiting a trade
opening price of the second bar to the right of the reverse bar
(Figure 25).
2. In case the bars did not cross the Zero line, but the slope is
reversed:
a. One should wait until the new bars exceed the highest or
lowest bar of the right shoulder (depending on whether
it is a Normal or Inverted Head-and-Shoulders pattern).
When a bar closes higher than the peak, or lower than the
bottom, of the right shoulder, the trade must be closed at
the opening price of the following bar to the right
(Figure 26).
b. Otherwise, the pattern may be forming a second right
shoulder that is shorter than the first right shoulder (i.e.
does not exceed the first right shoulder). In this case
the pattern should be treated as if it has only one right
shoulder (Figure 27). Figure 28: Re-entering a trade

Figure 25: Exiting a trade

Statistical Analysis of the Head-

and-Shoulders Pattern on the MACD
Histogram Indicator
Figure 26: Exiting a trade
Here the daily charts of Dow Jones Industrial Average (DJIA),
General Electric Stock (GE), Silver and British Pound Sterling
(GBP) for the period between January 1983 and December 2009
for conducting this statistical analysis are used.
During the specified period mentioned above, sixty-seven
Head-and-Shoulders patterns appeared on the MACD Histogram
indicator in all the four investment products under examination
in this analysis.
The pattern appeared 19 times on each one of the DJIA and
the GBP. In 79% of the DJIA’s cases the pattern was a success,
while the success rate for the GBP was 84%. On the charts of
the General Electric Stock, the pattern appeared 17 times and
succeeded in 82% of the cases. The lowest success rate was 75%
of the 12 cases that appeared on Silver charts.
Out of the total 67 appearances of the Head-and-Shoulders
pattern on the MACD Histogram indicator, the pattern was
successful in about 81% of the cases (54 times), while it failed
In some cases after exiting the trade, one can find the bars in 19% (13 times). This means, that the success rate of the
reversing again. In this case, we consider the new reverse a pattern is more than 4 times the rate of failure, indicating a very
second taller right shoulder, and we can re-enter the trade again, high statistical significance of the accuracy of the Head-and-
if we see the usual entry signal on the second right shoulder Shoulders pattern on the MACD Histogram indicator.
(Figure 28).

IFTA.ORG PAGE 13
 IFTA JOURNAL 2012 EDITION

Figure 29: Frequency distribution of the Head-and- The Normal form of the Head-and Shoulders pattern
Shoulders pattern on the MACD Histogram indicator for appeared 33 times, representing 49% of the total 67
each of the four investment products appearances of both forms of the pattern, while the remaining
51% represents the 34 incidents, in which it appeared in the
Inverted form.

Figure 33: Percentage of success/failure for the total

appearances of each of the two forms of the pattern

Figure 30: Success and failure cases of the Head-and-

Shoulders pattern on the MACD Histogram indicator for
each of the four investment products

In 24 out of 33 cases, the Normal form of the pattern was

profitable. The 88% success rate of the Inverted form of the
pattern presents 30 cases of success out of 34 appearances
in all of the four investment products. The largest portion of
successful patterns were seen in the DJIA at 100%, as 13 out
of 13 Inverted patterns founded on the DJIA were successful
cases (Figure 34). GBP had the most number of appearances
with a success rate of 85% of the 13 appearances of the Normal
patterns (Figure 35). Inverted Head-and-Shoulders pattern
Figure 31: Percentage of success/failure for the total proved to be the most successful form of the pattern and of very
appearances of the pattern high statistical significance.

Figure 34: Predictive frequency rate of the Inverted Head-

and-Shoulders pattern on MACD Histogram indicator for
each of the four investment products

Figure 32: The frequency of appearance of the Normal and

Inverted forms of the Head-and-Shoulders pattern on the
MACD Histogram indicator for each of the four investment
Figure 35: Rate of successful predictive frequency for the
products
Normal Head-and-Shoulders pattern on MACD Histogram
indicator for each of the four investment products

PAGE 14 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

Fifty nine percent of all successful trades in all the four Half of the successful trades in General Electric stock
investment products together fall in the profit bracket of 1% generated profits up to 3% and the other half more than 3% gain,
to 3%, when using the trading method explained earlier in this while 100% of the unsuccessful trades had a loss rate of less
paper, while the remaining 41% gains are more than 3%. Using than 0.6%. General Electric recorded its most successful trade
the same method, we find that the majority of losses are less during the period under examination in November 1989, when
than 1% with a rate of 77% of unsuccessful trades. Just one an Inverted Head-and-Shoulders pattern preceded a price of
case out of the 13 unsuccessful trades recorded a loss of 4.7% the stock to gain 12%, while the lowest profit percentage during
(in Silver), which is the highest loss rate recorded among all the same period was recorded two times in February 2007
cases. One out of the thirteen failures was a breakeven at the and August 2008 with a 1.2% percent profit. On average, and
specified exit point of the trade. Successful trades generated on following the pattern indicator, General Electric gained 4.2%,
average 4% profit per trade. The average loss per unsuccessful while it lost an average of 0.28% per unsuccessful occurrence of
trade was 0.91%, which is much lower than the average profit. the indicator. (Figure 39).
Approximately 2/3rds of successful trades generated more than
2% profit. The highest profit percentage was 23%, recorded in Figure 39: Distribution of profit percentages for General
Silver (Figures 36 and 37) Electric Stock

Figure 36: Distribution of profit percentages for the total

of all investment products

Throughout this statistical analysis, Silver proved to be the

most volatile among all the four investment products under
Figure 37: Distribution of loss percentages for the total of examination with profit percentages distributed among many
all investment products categories. Silver recorded the highest profit percentage per
trade among all the investment products when it gained 23%
profit in September 2008, as well as the greatest loss rate of
4.7% in November 1983. Average profit percentage for Silver is
7.9% per trade and average loss per trade is 1.9% (Figure 40).

Figure 40: Distribution of profit percentages for Silver

Highest profit percentage in DJIA was 12% and was recorded

in October 2002, while the greatest loss was 1% in April 1995.
DJIA average profit per trade is approximately 4%, while its
average loss is less than half percent. The lowest percentage
profit among all DJIA successful trades was 1.2%. Fifty four
percent of the successful trades fall between 1% and 3% profit
per trade (Figure 38). The British Pound Sterling (GBP) proved to be the most
interesting investment product under examination. It was
Figure 38: Distribution of profit percentages for Dow the most consistent in terms of profit/loss percentages with
Jones Industrial Average the least extremes as compared to the other three investment
products. Eighty eight percent of the successful trades
generated 1% to 3% out of which 65% fall between 1% and 2%.
Its highest profit percentage, after appearance of the indicative
pattern, was in April 1985 with a rate of 4.8%, while lowest profit
rate was 1%, and was recorded in seven cases. Its worst loss was
1.9% in June 1988 and the average loss per trade was 1.2% in only
three unsuccessful trades. On the other hand the average profit
in 16 winning trades was 1.87% per trade (Figure 41).

IFTA.ORG PAGE 15
 IFTA JOURNAL 2012 EDITION

Figure 41: Distribution of profit percentages for British Pound Conclusion

Sterling This paper—supported by the results of the statistical
analysis—demonstrated that the Head-and-Shoulders
patterns appearing on MACD Histogram indicator can be
used to make long or short trading decisions in any liquid
market or investment product. Indicators are usually used as
supporting tools to confirm a trade entry and/or exit, but this
paper demonstrated, that the MACD Histogram indicator and
associated Head and Shoulders pattern formation can also be
used to help allocate a specific entry/exit point of a trade.
Duration of trading the Head-and-Shoulders pattern on the The statistical analysis conducted strongly indicates that
MACD Histogram indicator from entering the trade position the methodology applied to trading offers an expected risk to
until exiting it is an additional and important consideration. reward ratio of 4.4. In addition to the very good risk to reward
The average duration per trade was around eight vertical bars of ratio, the analysis showed that failures counted to less than
the MACD Histogram bars, which represents eight days in this 20% of total frequency of appearance of the pattern, which also
statistical analysis. As we know each bar of the MACD Histogram represents only 24% of the 54 winning cases. This means, that
indicator represents one bar on the price side of the chart. This each time the pattern is traded the expected rate of success,
price-side bar can represent a minute, an hour, a day, a week or a including the percentage profit, is statistically significant.
month depending on the time-frame used for the chart. On the daily period under statistical analysis, the pattern
In figure 47, we see that almost 70% of the trades lasted proved to generate acceptable average profit percentage per
between seven and 10 days, with the highest rate of occurrence for trade in a relatively short period of time under consideration,
the seven days duration, representing 26% of the total number of which amounted to an average of eight days per trade. This is
successful trades. Second to this comes the 10 days duration with considered acceptable duration relative to the profit outcome.
18% of all successful trades. Each of the four investment products, The Head and Shoulders pattern as applied to the MACD may
separately, ranged between eight and 8.6 days of duration on be used as a predictive trading indicator to the long or short side
average, which is around the average duration mentioned above. in liquid markets and under various time frames.
Five days was the lowest number needed to complete a winning
trade for the period and investment products under examination. References
Murphy, John J., Technical Analysis of the Financial Markets, NYIF, 1999.
There was no trade found to last less than five days or more than Pg. 255
14 days. One case in each investment product lasted five days, Elder, Alexander, Trading for a living, Wiley, 1993. Pg. 102
while the 14 days duration was recorded only twice, once in DJIA
and the other in General Electric Stock (Figure 42). Bibliography
Edward Robert D. and John Magee, Technical Analysis of Stock Trends, 7th
Edition, Amacom, 1997
Figure 42: Percentage of occurrence for duration per Elder, Alexander, Trading for a living, Wiley, 1993
successful trade Murphy, John J., Technical Analysis of the Financial Markets, NYIF, 1999
Pring, Martin J., Technical Analysis Explained, McGraw Hill, 1991

Software and data

Data and charts provided by Reuters
Dow Jones chart example of the year 1907 provided by Prophet.net

Endnotes
1 Murphy, John J., Technical Analysis of the Financial Markets, NYIF, 1999. Pg. 255
2 Elder, Alexander, Trading for a living, Wiley, 1993. Pg. 102

If this methodology is applied on weekly charts, it is expected

to generate greater gains in percent profit. Larger percentage
profit require longer durations and when considering weekly
charts, each vertical bars of the MACD Histogram indicator
represents one week in this statistical analysis.

Risk /Reward Ratio

According to the methodology of trading the Head-and-
Shoulders pattern on MACD Histogram indicator and to the
findings of the statistical analysis, if we take the average gain of
4% per trade and divide it by the average loss of 0.91%, we have a
positive risk/reward ratio of 4.4.

PAGE 16 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

Trading Strategies Based on New Fluctuation Tests

by Daniel Ziggel ([email protected]) and Dominik Wied ([email protected])

Abstract derive trading strategies. Moreover, we analyze the dates of

There are many empirical hints that correlations and rejection and the resulting parameter estimators. It turns out
variances among many time series cannot be assumed to that the tests perform very well throughout the whole empirical
remain constant over longer periods of time. In particular, application and the resulting dates of rejection seem to be
correlations and variances among stock returns seem to reasonable. Additionally, the resulting trading strategies seem
increase in times of crisis. An increase of correlations and to be promising.
volatilities has serious consequences for diversification
effects, which lie at the heart of several applications in risk Model and Test Statistics
management and portfolio optimization. In this paper, we Let be the time series of an asset. For
analyze recently proposed tests to determine if correlations example, X t  might be the final quote of day t. We want to
and variances of assets are constant over time. An empirical determine whether the variance of X t is constant over time, i.e.
application to various assets and trading strategies suggests we test:
that the tests perform well in practice.

Introduction Here, σ 2 is assumed to be constant. The test statistic is given by:
During the recent financial crisis, capital market volatilities
and correlations increased quite dramatically. As a consequence,
risk figures increased significantly, diversification effects were
overestimated and ultimately, capital was lost. In literature, The test rejects the null hypothesis of constant variance
this phenomenon is sometimes referred to as “Diversification if the empirical variance fluctuates too much. The
Meltdown” (Campbell et al., 2008). Moreover, there is a fluctuation is measured by . The
consensus in empirical finance, that market parameters cannot weighting factor scales down deviations at the beginning,
be assumed to remain constant over longer periods of time (e.g. where the fluctuations are more variable. Besides, the scalar
Krishan et al., 2009). In particular, the relevant parameters seem factor is required to derive an asymptotic distribution under
to increase in times of crisis. A comparison of correlations and the null and is cumbersome to write down, but can easily be
volatilities during different market phases in the last ten years calculated from the data. The complete formula can be found in
can be found in Bissantz et al. (2011a, 2011b). Wied et al. (2011a).
A diversification meltdown has serious consequences for Under some mild theoretical conditions, the asymptotic null
applications in finance. For example, portfolio optimizations distribution of the test statistic is a one-dimensional Brownian
which are based on diversification effects between several bridge. This distribution is well known (Billingsley, 1968). Using
assets are no longer valid if the parameters changed. Similar the quantiles of this distribution, we obtain an asymptotic test
problems occur with applications in risk management or for for our problem.
the valuation of financial instruments. Surprisingly, there is a The test for constant correlations is quite similar.
lack of methods to formally test for changes in correlations or Nevertheless, we need two time series and
volatilities as most existing procedures either require strong . With this, the test statistic is given by:
parametric assumptions (Dias and Embrechts, 2004), assume
that potential break points are known (e.g. Jennrich, 1970), or
simply estimate correlations from moving windows without
giving a formal decision rule (e.g. Longin and Solnik, 1995). Only The expression is the empirical correlation coefficient
recently, Galeano and Peña (2007) and Aue et al. (2009) have calculated from the first observations. The test rejects the null
proposed formal tests for a change in covariance structure, hypothesis of constant correlation if the empirical correlations
which do not build upon prior knowledge as to the timing of fluctuate too much, as measured by .
potential shifts. The test of Aue et al. is based on cumulated Again, the weighting factor scales down deviations at the
sums of second order empirical cross moments (Ploberger et al., beginning, where the fluctuations are more variable. Moreover,
1989) and rejects the null of a constant covariance structure if the scalar factor captures the volatilities of and as well
these cumulated sums fluctuate too much, while Galeano and as the dependence of over time in order to derive the
Peña work in a parametric environment. asymptotic null distribution. As before, the asymptotic null
In this paper, we use tests proposed by Wied et al. (2011a, distribution of the test statistic is a one-dimensional Brownian
2011b) which focus on correlations and volatilities in order to bridge. The complete formula can be found in Wied et al. (2011b).

IFTA.ORG PAGE 17
 IFTA JOURNAL 2012 EDITION

Test Procedure The lower the significance level is chosen, the earlier and more
As mentioned above, quantiles of the asymptotic distribution sensitive the test reacts and the other way round. Our empirical
can be found in related reference books. Popular quantiles are results suggest that α=5% is a reasonable choice for a lot of
given by: applications.
Besides, large differences of the market parameters between
ƒƒ 1.073 (80%) the break points can be observed hinting at a reasonable
ƒƒ 1.224 (90%) separation of different market phases. This phenomenon provides
ƒƒ 1.358 (95%) the basis for our trading strategy described in the next section.
ƒƒ 1.628 (99%) Table 3 illustrates this phenomenon for the DAX and shows the
annualized market parameters (returns and volatilities) for the
For example, if , we can reject the null of respective period between two structural breaks.
constant correlation on the significance level 5%. Roughly
spoken, the correlation has changed with a probability of 95 Table 1: Structural breaks (Volatilities, α=5%)
%. Moreover, if we evaluate the term on its S&P DAX REX CRB
own, we can appraise if the correlation increased or decreased. 03.02.1988 29.01.1988 12.02.1990 15.03.1988
If holds for the maximum of the test statistic, the 03.02.1989 28.10.1988 06.06.1994 03.06.1988
correlation increased and also the other way round. Of course, 12.02.1993 28.11.1988 03.04.1995 07.09.1988
this information is important for several applications. The same 30.09.1993 01.02.1989 01.05.1995 02.02.1990
principle applies for . 18.07.1996 19.04.1989 17.11.1995 09.11.1992
10.03.1997 07.06.1989 19.02.1996 05.08.1994
Structural Breaks 02.02.2005 20.05.1993 28.07.1998 02.12.1996
In order to evaluate the quality in applications, the tests 19.10.2007 17.06.1993 12.12.2000 07.01.1998
are applied to several time series of assets: two stock indices 28.10.2008 14.03.1994 14.05.2001 19.05.1998
(S&P 500, DAX), a commodity index (CRB Spot Index) and a 09.10.2009 19.08.1997 01.08.2003 18.06.2001
government bond index (REX), using daily data (final quote) and 27.11.2002 18.10.2004 19.10.2001
a time span of 22 years (January 1988 to April 2010). 07.04.2003 29.02.2008 23.05.2003
The procedure for the test is as follows. We start at the 17.09.2003 27.01.2010 23.06.2003
20th available data point and increase the period of time 06.02.2004 21.07.2003
successively for one day. The starting point is due to the fact 07.03.2005 10.05.2004
that approximately 20 data points are required for a reliable 14.07.2006 02.09.2008
estimation of the respective parameter. This procedure is 06.10.2006
performed until the test rejects the null hypothesis of constant 14.03.2007
correlation resp. variance. Then, the 20th day after rejection 24.07.2007
is the new starting point and the procedure is repeated for 06.11.2007
the remaining time span. This procedure is due to the fact 24.11.2008
that the parameter can no longer be assumed to be constant, 28.08.2009
if the null hypothesis is rejected. A new
reliable estimation requires once again 20
data points after the point in time, where Table 2: Structural breaks (Correlations, α=5%)
the parameter changed. Otherwise, the
S&P & DAX S&P & REX S&P & CRB CRB & DAX CRB & REX DAX & REX
estimator would be biased as data of two
different phases were mixed. 11.02.1965 04.08.1998 25.09.1981 17.07.1981 16.06.1988 13.11.1989
Tables 1 and 2 include the rejection dates 28.06.1965 01.09.1998 14.12.1981 10.10.1986 18.07.1988 11.12.1989
of the null hypothesis for the significance 13.05.1970 31.01.2000 11.01.1982 21.10.1987 15.08.1988 08.01.1990
level α=5%. The results seem to be 22.10.1987 08.03.2000 05.03.1985 24.02.1999 31.01.1989 29.10.1997
reasonable. Moreover, there is a strong 23.12.1999 22.12.2000 26.10.1987 25.03.2002 01.03.1989 05.03.1998
dependence between structural breaks and 20.01.2000 28.08.2002 11.02.1999 22.04.2002 09.09.1998 07.04.1998
distinctive changes in trends. For example, 22.11.2000 15.10.2002 11.03.1999 28.06.2002 23.09.2008 05.05.1998
there are a lot of rejections between 2000 20.12.2000 01.08.2003 09.10.2008 17.03.2008 15.06.1998
and 2003 (Dotcom-crisis) and in 2008 10.04.2001 11.04.2008 07.07.2008 21.08.1998
(financial crisis). In contrast to that, there 14.09.2001 30.09.2008 04.08.2008 18.09.1998
are only a few structural breaks in stable 21.10.2002 28.10.2008 01.09.2008 16.10.1998
market phases. 10.12.2002 10.10.2008 14.06.2002
Our results show that the chosen 07.01.2003 01.08.2003
significance level plays an important role for 25.03.2003 08.09.2008
both rejection frequency and rejection dates. 22.02.2008 14.10.2008
Consequently, the significance level has to 15.10.2008 11.11.2008
be chosen carefully in practical applications. 09.12.2008

PAGE 18 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

Table 3: Rejection dates and annualized market parameters

(Volatilities, α=5%) test, we perform an out of sample study. In this study, we
investigate a simple strategy which applies the proposed test.
DAX Returns Volatilities The strategy is as follows. The available time span since
29.01.1988 - 28.10.1988 44.09% 16.07% the last detected change in volatility is used to calculate the
28.10.1988 - 28.11.1988 -34.9% 12.63% historical return, which is used as an estimator for the future.
28.11.1988 - 01.02.1989 15.34% 13.29% Moreover, an asset is allowed to be bought if the last structural
01.02.1989 - 19.04.1989 27.35% 12.84% break lies 20 days or more in the past and the return estimator
19.04.1989 - 07.06.1989 9.99% 8.96% is positive. A short position is opened, if the last structural
07.06.1989 - 20.05.1993 3.26% 20.04% break lies 20 days or more in the past and the return estimator
20.05.1993 - 17.06.1993 54.55% 7.87% is negative. Finally, the capital is uniformly allocated between
17.06.1993 - 14.03.1994 28.53% 16.41% all allowed assets.
14.03.1994 - 19.08.1997 18.48% 14.59% In order to guarantee an objective back-test, we assume the
19.08.1997 - 27.11.2002 -4.46% 28.87% following:
27.11.2002 - 07.04.2003 -49.58% 43.45%
07.04.2003 - 17.09.2003 63.03% 25.92% ƒƒ The test is performed daily (final quote).
17.09.2003 - 06.02.2004 29.14% 19.17% ƒƒ Portfolio shifting is done the next day (final quote).
06.02.2004 - 07.03.2005 8.63% 15.08% ƒƒ We neglect transaction cost, taxes, fees and currency
07.03.2005 - 14.07.2006 15.73% 14.47% fluctuations.
14.07.2006 - 06.10.2006 39.39% 14.67% ƒƒ The rebalancing is performed daily.
06.10.2006 - 14.03.2007 19.13% 11.66% ƒƒ The significance level is α=5%.
14.03.2007 - 24.07.2007 48.34% 15.69%
24.07.2007 - 06.11.2007 -5.78% 16.18% Note that the strategy is based on indices, which can be
06.11.2007 - 24.11.2008 -58.16% 33.42% implemented by means of ETFs and ETCs. Hence, transaction
24.11.2008 - 28.08.2009 35.39% 33.66% costs play only a marginal role. We use the following indices for
28.08.2009 - 01.04.2010 21.25% 18.55% our strategy:

ƒƒ EuroStoxx 50 (Long & Short)

Application Areas ƒƒ MSCI Emerging Markets (Long)
Before describing our trading strategy in detail, we want to ƒƒ Gold - troy ounce (Long)
present several application areas of the tests. When developing ƒƒ iBoxx euro zone 3-5 Years (Long & Short)
the tests, we first wanted to find optimal time points for the
re-optimization of a portfolio, i.e. we wanted to use the The choice is due to two reasons. On the one hand, the
tests for an optimal timing. A re-optimization is necessary investment universe should be simple and clearly arranged. On
if the input-parameters (correlations or volatilities) change the other hand, there should be different instruments available
significantly so that parameter estimation and portfolio in order to react to various market phases. This is guaranteed
optimization based on data points before the structural break by means of six allowed trading options. These options cover
would then no longer be valid. increasing and decreasing stock markets and interest rates.
Another potential application is the implementation of Moreover, commodities protect against rising inflation. Hence, a
an alert-function. For example, if an unfavorable parameter lot of different scenarios can be covered. Finally, all indices can
changes occur, warnings for risk management are generated. be implemented by means of ETFs and ETCs.
With the help of such an alert-function, also estimation of
measures like the value-at-risk might be improved. In addition, Results
as explained previously, the operator can detect trends, which We will only present results for the trading strategy, which is
might be visualized by turning lights. Importantly, whole based on structural breaks of volatilities as the results are quite
trading strategies basing on structural breaks can be developed. similar for the strategy based on correlations. The results can be
found in Table 4 and Figure 1.
Trading Strategy The strategy yields an above-average return in comparison
The results above show that changes in market parameters to the underlying indices. Moreover, the portfolio development
can reasonable be detected. In order to investigate the possibility is very smooth—even throughout financial crisis. This result is
to derive trading strategies, which are based on the proposed remarkable as three risky assets are used within the strategy.

Table 4: Results of the trading strategy and summary statistics

Strategy EuroStoxx 50 MSCI EM Gold iboxx euro zone Average

Return p.a. 10.16% -0.15% 10.45% 12.77% 4.01% 6.77%
Volatility p.a. 10.02% 22.47% 17.24% 17.99% 2.59% 15.07%

IFTA.ORG PAGE 19
 IFTA JOURNAL 2012 EDITION

Compared to average values there is a significant improvement finance and have found that the tests perform well in these
of the relevant parameters. More precisely, the return increases applications. Moreover, we have derived a simple trading
by 50.07% whereas the volatility decreases by 33.51%. The strategy and have proved its usefulness by means of an out-
Sharpe Ratio is about 1—also a good value. of-sample study. For sake of simplicity, we have tried to avoid
Figure 2 shows the development of portfolio weights over mathematical details and cumbersome formulas.
time. Between crises, the weights are very stable whereas in Nevertheless, some fundamental questions still remain: It is
times of crises a lot of fluctuations can be observed. This fact of interest to specify the economic determinants of fluctuations
guarantees the good performance and the prevention of (high) to model and forecast variations of correlations and variances.
losses during financial crisis. And the question arises if parameter estimators, based on
the new tests, will improve the performance of portfolio
optimization. Portfolio optimization depends, of course, on a
Figure 1 reliable estimation of market parameters. These topics will be in
focus of our ongoing research.

Acknowledgements
Financial support by Deutsche Forschungsgemeinschaft
(SFB 823, Statistik nichtlinearer dynamischer Prozesse, project
A1) is gratefully acknowledged.

References
1. A. Aue, S. Hörmann, L. Horváth, M. Reimherr (2009): “Break detection
in the covariance structure of multivariate time series models”,
Annals of Statistics, 37(6B), 4046-4087.
2. P. Billingsley (1968): “Convergence of probability measures”, Wiley,
New York.
3. K. Bissantz, N. Bissantz, D. Ziggel (2011a): “An empirical study of
correlation and volatility changes of stock indices and their impact on
risk figures”, Acta Universitatis Danubius (Œconomica), forthcoming.
4. N. Bissantz, V. Steinorth, D. Ziggel (2011b): “Stabilität von
Diversifikationseffekten im Markowitz-Modell”, AStA Wirtschafts-
und Sozialstatistisches Archiv, Vol. 5, Nr. 2, 145-157.
5. R. Campbell, C. Forbes, K. Koedijk, P. Kofman (2008): “Increasing
Figure 2 correlations or just fat tails?”, Journal of Empirical Finance, 15, 287-
309.
6. A. Dias, P. Embrechts (2004): “Change point analysis for dependence
structures in finance and insurance”, Risk measures of the 21th
century, Wiley, New York.
7. P. Galeano, D. Peña (2007): “Covariance changes detection in
multivariate time series”, Journal of Statistical Planning and
Inference, 137(1), 194-211.
8. R. I. Jennrich (1970): “An asymptotic chi-square test for the equality
of two correlation matrices”, Journal of the American Statistical
Association, 65, 904-912.
9. C.N.V. Krishan, R. Petkova, P. Ritchken (2009): “Correlation risk”,
Journal of Empirical Finance, 16, 353-367.
10. F. Longin, B. Solnik (1995): “Is the correlation in international equity
returns constant: 1960-1990?”, International Money and Finance,
14(1), 3-26.
11. W. Ploberger, W. Krämer, K. Kontrus (1989): “A new test for structural
stability in the linear regression model”, Journal of Econometrics, 40,
307-318.
12. D. Wied, M. Arnold, N. Bissantz, D. Ziggel (2011a): “A new fluctuation
test for constant variances with applications to finance”, Metrika,
forthcoming.
We want to point out that alternative and more sophisticated 13. D. Wied, W. Krämer, H. Dehling (2011b): “Testing for a change in
strategies can easily be constructed by means of the fluctuation correlation at an unknown point in time”, Econometric Theory,
tests. However, we consciously chose this simple and traceable forthcoming.

strategy in order to demonstrate the principle and benefit. As

this simple strategy yields already remarkable results, we expect
a refinement of the strategy to improve the results even more.

Conclusion and Outlook

In this paper, we have described two new tests to determine
whether correlations and variances of time series are constant
over time and have investigated their performance in several
applications. To this end, we have applied the tests to several
time series of assets which are relevant for applications in

PAGE 20 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

An Introduction to W. D. Gann’s Time Cycles

by Mikael Bondesson

Abstract Figure 2
A short review of the principles behind W.D. Gann’s work on
Time Cycles and examination of the phenomenon seen behind
these cycles on three equities, including Skandia, Ericsson,
Astra Zeneca and the Nasdaq100

Figure 1
Introduction
This paper will attempt to
demonstrate an implausible
phenomena: one that moves
360 Days us to contemplate about
how precisely the markets
do, indeed, vibrate through
time. Price
This will be my
interpretation of one aspect
0 I have studied on W D Gann’s
large work, where the major
sources are what I have
learned from my Swedish
teacher Ingemar Carlsson,
who showed me this form 45 90 135 180 225 270 360 405 450 Time
180 of analysis for the first time
in 1998, and my Australian/
American teacher Bill The Yearly Cycle
0 McLaren, of whom I have As a fresh student on Gann’s time factor, I think the Yearly
been a student since 2000. Cycle is the best time consideration to start one’s studies. This
Both of these gentlemen is because we all can relate to the year as a cycle and then the
90 270 are still my biggest source seasons within that particular cycle. Additionally the yearly
of inspiration and new cycle presents common phenomenon for many of the world’s
insights. indices, stocks, currencies etc.
180 Please also note that A year consists of 365 calendar days, but to give our analysis
Gann Analysis is just one some margins of order and consistency, we shall define every
of several tools I use for month as 30 days, and therefore the year in this analysis
0 technical analysis. I have contains only 360 days. As we all know, a circle consists of 360
45 315
found the pattern of trends degrees, so now we can look at the Yearly Cycle as a circle with
and what actually is on the its 360 degrees (days) in its full revolution. This method will
90 270 chart, to be the basic and make our study more instructive.
most important element In both his time and price analysis, Gann divides the cycles
135 225 to construct an analysis of and movements into 1/8´s and 1/3´s and their multiples and
180 a stock. To this we can add divisions, to determine the divisions of the cycles. See figure 1,
volume analysis, indicators, and note the division of the Yearly Cycle: Dividing 360 days by 2,
intra-market analysis, wave gives us 180 days, and dividing this again by 2 gives us 90 days,
theory, Fibonacci, etc., and time analysis. which of course also is 360 days divided by 4. If we again divide
This text does not claim to be a full description of W D by 2 we get 45 days, which as we know is 1/8 of 360 days.
Gann’s time cycles. Instead, my intention is to give the reader Dividing 360 days into 1/3´s gives us 120, 240 and 360 days
some insight to the interest and value of Gann Theory, with a in the yearly count. And from the 1/8 and 1/3 divisions together,
brief theoretical introduction and a few thorough examples to we can derive the divisions by months, or 30 days, 60 days and
hopefully evoke a curiosity and desire for further studies. 90 days etc. We now make a division in the circle so that we can

IFTA.ORG PAGE 21
 IFTA JOURNAL 2012 EDITION

place it along the x- or time axis on the chart as shown in figure 2. Skandia
Continuing in this analysis, we then count the days from Let’s turn to figure 4, the Swedish insurance company
determined significant highs and lows to identify if the Skandia, during 2001-2004 (since acquired by Old Mutual).
particular cycle at hand is valid. However, we do not expect At a first glance there doesn’t appear to be a particular cyclical
to see exact reactions at every single date we think might be pattern on this chart, but instead more an irregular chopping
important. Instead, we “Use them as a road map to measure the up and down price action. If we zoom in the development from
duration of moves and also as general time periods to look for the significant high in February of 2004 where Skandia turned
more significant changes in direction”, as McLaren expresses it into a creeping trend to the downside, and look at that a bit more
describing the day counts. closely, we find a very precise pattern.
For example, suppose a stock has been trending down
for many months, however in recent time with decreasing Figure 4
momentum (see figure 3). In our illustration, on the 21st of
December the stock reaches a low and starts to trade up for
three weeks before declining again, but this time reaching a
higher low on the 21st of January. That the stock has made a
higher low is as we all know and possible sign that it could be
going from the falling trend to something else (in addition to the
stock breaking the falling trend line). But because it is exactly
30 days between these two lows, we now also have a sign that
the stock might have started to vibrate with the Yearly Cycle.

Figure 3

180

150
We will allow +/- 1 day at the dates for the cycle turning
135
points, to give the analysis margin. This is one method of various
techniques. As we are working with calendar days it follows that
30
90 if the expected turning point will occur at a weekend, we must
0 allow the turning point to occur during the trading sessions both
at the Friday before or the Monday after that particular date.
21 21 21 6 21 21
12 1 3 5 5 6 The following weeks of trading can then give a clue on which
date actually was the date from a cyclic perspective.

Figure 5
During the following two months, the stock trades up and
down in a base pattern, making yet another higher low on the
21st of March, exactly 90 days after the first significant low. Such
an observation would be additional evidence that the stock has
started to vibrate with the Yearly Cycle. Then breaking obvious
resistance, the stock takes off in a rally. After a period of time,
in this illustration, it then reverses into 135 and 150 days from
low, indicating further gains on the upside, after which the stock
exhausts up into 180 days from low on the 21st of June, signalling
a possible high or at least significant resistance in time. This was
one example and illustration of how time cycles could be set-up
based on the above price progression theory.
Does this work in practice? To explore and answer this
question we will review examples from the financial markets,
most of which I have followed in real time as they unfolded
through time. The following examples have been prepared
for various lectures held at the University Of Lund, Sweden,
and other venues over the last years. Added to this is a recent
example observed in the Nasdaq-100, which the reader can Viewing figure 5, we note that Skandia reached an important
continue to monitor from the analysis given here. high on the 13th of February 2004, after which it went into a
creeping downward trend. This day was a Friday, so the cyclical

PAGE 22 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

top could have come from the Thursday before, to the Monday in mid August, which we identified earlier. As we see, this low is
thereafter. As we see by the chart, the following months of 45 days from high, 90 days from low, 135 days from high and 180
trading would, however, suggest that the cyclical top came in days from the top. We have observed here the accuracy of the 45
at the Monday, on the 16th of February. We’ll soon discover why day cycle within the 90 and Yearly Cycle.
this is so. As we see by the chart, the next move up from the 16th of
Observing three months later, we find that a low was August lasted another 45 days, bringing in another lower high on
established in mid May, or more precisely on the 17th of May Monday the 4th of October at the falling trend line. That Monday
(Monday), indicating a cyclical low between the 14th (Friday) and is within the defined time window, Friday/Monday, and as we
18th (Tuesday). This is approximately 90 days after the top in see, it produced the high for another run down and a double
February, so a possible Yearly Cycle present. bottom some three weeks later in late October. Another 45 days
Another three months later a new significant low was later, and 90 days from the low in August, yet another lower high
established in mid August, on Monday the 16th to be exact. The was established within the falling trend line and resistance, on
cyclical low should be between the 13th (Friday) and the 17th the 15th of November. At this point however, pattern of trend had
(Tuesday) of August, or 90 days from the May low, and of course started to indicate to us a possible change in trend which was
180 days from the February high. The stock traded 90 days high confirmed by the break of the trend line a few weeks later. This
to low, and then 90 days low to low. Looking 90 days ahead from started a new uptrend, but the 45 day cycle continued to show
that time period, we find that Skandia reached a new high on some presence for yet some time. Look another 45 days ahead,
Monday the 15th of November. The trend was now once again and see how it brought in and fits an Elliott wave three high and
about to change to an upward trend after 10 months, so not a reversal for two weeks.
large reaction in price this time. Now go back to figure 4 on Skandia and we can see and
But let’s go back to study these 10 months in more detail. conclude that in what seemed to be nothing more than an
So far we notice how a very precise 90 day cycle is marked by irregular pattern, was instead a very precise 45 day cycle
important highs and lows during this time period. But there are present.
other significant turning points on the chart. Are they somehow
linked to this rhythm? Ericsson
Half of the 90 day cycle is 45 days, which is 1/8 of 360 days. Let’s turn back the clock … and look at Ericsson, a company
Having discovered a very accurate 90 day cycle, let’s continue that has been of great economic importance in Sweden. During
our research to see if we can find 45 day increments within the euphoric era of that late 1990s to 2000, and after having
those 90 day vibrations. Adding 45 days from the top on the 16th performed very well for a number of consecutive years,
of February gives us early April as a probable turning point. The Ericsson’s stock, which was the most heavily traded equity on
chart confirms our theory (figure 6). The stock has made a lower the Stockholm Stock Exchange, exhausted up to it’s final top in
high on the 5th of April which was a Monday; and we conclude March of 2000 (figure 7).
that Friday the 2nd of April in fact is the cyclical turning point.
Indeed, Skandia then falls another 45 days in to the 90 day Figure 7
bottom on the 17th of May as can be seen on the chart.

Figure 6

After reaching a top and all-time high at 117 Swedish crowns,

on Monday the 6th of March 2000, Ericsson went into a topping
From that date the stock starts to rise for 45 days again, formation for over six months, after which it started a severe
into another lower high on Friday the 2nd of July. Note that it decline reaching a final bottom in September of 2002 at 2.96
follows that this is then 90 days from the first lower high at the crowns(!!), at which point there was doubt about the company’s
2nd of April, and 135 days from the top price where this moved ability to survive. The final top and all-time high is obviously a
initiated. Naturally, there is then 45 days to the next 90 day low very significant top for Ericsson, so let’s begin from there and

IFTA.ORG PAGE 23
 IFTA JOURNAL 2012 EDITION

see if we can find a Yearly Cycle. This lower high after 180 days is interesting because 180 is
a strong vibration point within the Yearly cycle. Here, Ericsson
Figure 8 gives a signal of another reversal move down. But looking at the
pattern of trend, this could also be a possible end to a topping
formation and the start of a downtrend, after having traded one
180-day and two 90-day cycle blocks, producing lower highs at
their expirations. This is then confirmed by the break of obvious
support in late September as seen in figure 9.
As an aside: In a downtrend, a counter trend reversal up into
an important time period indicates resistance in time and price,
with possible more to come on the downside. Likewise, in an
uptrend, a counter trend reversal down into an important time
period signals support in time and price, and that further gains
may be expected.
With Ericsson, and from the top in early September the
stock falls quite dramatically but reverses up in early October,
producing a lower high on the 5th (e.g. 30, 120 and 210 days from
high) after which it starts to fall again. In the second half of
October Ericsson begins its reversal again——producing another
Monday the 6th was the top, so either of the days from Friday lower high on Friday the 3rd of November (time window 60 days
the 3rd thru Tuesday the 7th, is the possible start of a new cycle from September the 5th). Ericsson continues to fall, repeating
(figure 8). Going 30 days ahead to early April; we see reactions the price action with a reversal up at the end of the month
in the stock both at the 3rd (high) and the 5th (low). Another 15 producing yet another lower high on December the 5th.
days later looking for a possible 45 day reaction, we find a new Once again, this is 90 days from the September high, 180
low on Monday the 17th of April. It is maybe a bit early but still days from the June high and 270 days from the top in March,
interesting to note at this point. This proves to be the start of a indicating resistance in time and that the downtrend could
new two week 36 percent rise up to two lower highs on the 2nd continue. Note also the reversal up to the 17th of October,
and 5th of May (Tuesday after holiday and Friday), that is 60 days 90  days from high and 180 days from low.
from the top in March. At the beginning of June, or 90 days from Looking another 90 days further (figure 10) we see the stock
the top, Ericsson reaches another lower high on the 5th (Monday), reversing up into the 4th of January (e.g. 30 and 300 days), Friday
which is then broken by marginally higher prices on the 8th (false the 19th of January (e.g. 45 and 315 days) and the 6th of March
break), followed by a 20 percent drop the rest of the month. The (e.g. 90 and 360 days). This high in March is exactly one year
120 day period from the high produces no significant turning from the top and completes the Yearly Cycle. This cycle continues,
point. But then on the 17th of July (Monday), 90 days from the however, to influence price reactions in Ericsson, although
April low we get another high, followed by a fast drop down until weakened and not as exact as the earlier cycle phase identified.
the 3rd of August, or 150 days from high. The stock then continues
to trade around that low and is held down both on the 4th (Friday) Figure 10
and the 7th (Monday) developing a short term base pattern, from
where it takes off up again and rises 26 percent in 30 days up to
180 days from high on the 5th of September (figure 9).

Figure 9

I have marked some dates and turning points in figure 10 and

11, for the reader to examine, but if we just take a look at how
the 90 day vibration continues to vibrate, we determine small
reversals up to the 5th and 6th of June and the 4th of September

PAGE 24 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

2001, but not decisive and questionable from an analytical or going to report the 1999 year result on that very day. I emailed
practical trading perspective. Ingemar to see if he had seen the setup with the 120 day move
On the 6th of December 2001, however, Ericsson makes a up from August to November and then the 30, 45 and 60 day
very significant top again, followed by a continued downtrend, increments on the way down from the November top.
a clear reversal in late February and a lower high on the 7th of
March 2002, two years from the top (with the time window Figure 11
extended to +/-2 days), indicating possible further declines. The
downtrend proceeds, however, with no significant reaction in
June during this cycle phase. In early September another lower
high is reached on the 6th at around 7 Swedish crowns, followed
by a final decline to the ultimate bottom at 2.96 crowns on the
30th of September 2002.
We have followed Ericsson down through the severe decline
experienced during from 2000 – 2002; and observed how
accurate the Gann Yearly Cycle vibrated through time, and the
signals generated. Ericsson is one of the largest and most traded
stocks at the Stockholm Stock Exchange, and every time I go
through this example I ask myself how much of the capital, or
better said the people behind it, are and were actually aware of
these cycles. My guess is very, very few.

AstraZeneca in 1999 - 2000

In the next example, AstraZeneca, we will look at a time
period where the Yearly Cycle was in force. AstraZeneca is a Figure 12
global biopharmaceutical company.
Ingemar Carlsson, mentioned in the introduction, showed me
this form of analysis for the first time in 1998, so back in 1999
I had just started experimenting with these cycles trying to
understand what to look for and what to expect from different
setups. At this time, I was closely watching AstraZeneca
(figure 12) trending up from the bottom on the 26th of July 1999
(Monday) reaching a temporary top on the 25th of August, and
nearly repeating that pattern of a temporary high (resistance
in time) a month later, or 60 days from low. The counts where
not exact and I struggled here, trying to find indications of the
presence of a possible Yearly Cycle. I anyway decided to monitor
the stock more closely the coming months. The high on the 21st
of October made me even more confused without a significant
high or low around the 24th – 27th in October as expected. Later,
I realized that AstraZeneca had produced a significant top 120
days from the low on the 24th of November 1999, and had started
to trend down again. With this, I understood that a cyclical
pattern could be unfolding and deserving of attention. “Yes, I’m monitoring this myself, doesn’t it look beautiful? It
Just before Christmas on the 22nd and 23rd of December, and sure looks as though AstraZeneca is going down into that date to
30 days from the high, Astra Zeneca produced a temporary low produce a low, but as I have told you before we must wait until the
(support in time), and then a lower high on the 10th of January cycle’s turning point to see if we really get a low established. As you
2000 (Monday) or 45 days from high. Later in January another know by now, there is also a chance we will get another lower high
temporary low came in 60 days from high on the 25th of January, and a continuation of the downtrend. The best thing would be if
followed by a reversal and then a subsequent continuation of the the first reaction to the report is negative, this should then exhaust
downtrend. the move down and wash out the sellers. But we’ll see.”
In mid February AstraZeneca was still trending down and The chart illustrates this is exactly what happened. The first
approaching the levels to around 290 SEK where it had bottomed reaction to the report was very negative and the stock traded
seven months earlier before it started to trade up for 120 days. down about 10% that day and closed near to the low at 266 SEK.
The 90 day time window around the 24th of February was going The next day AstraZeneca immediately reversed and started
to be an important date from a cyclical perspective. If it provided a rapid uptrend. I rang Ingemar and we both laughed at the
a low, it could mark a possible change in trend, leading to a more outcome. And he continued:
significant move up. So my excitement grew day by day as the “And did you see where it bottomed? At 266 SEK!”
stock continued down. In addition to this, the company was “Eeeeh… yes…?” I said

IFTA.ORG PAGE 25
 IFTA JOURNAL 2012 EDITION

“135 SEK from the top at 401!!! The stock trended down 135 SEK Nasdaq-100 in 2010 - 2011
in 90 days. Time and price was in harmony! Magnificent!” To update this study in mid June 2011, I also want to add a
Astra Zeneca then rose rapidly for 30 days (figure 13), to more recent example from the current markets. The Yearly
a temporary top on the 24th of March before correcting to Cycle has been vibrating in many markets this year and to
continue up to another temporary top on the 25th of April. This illustrate this, we will look at the Nasdaq-100 index, where it is
was the first day the markets were open again after Easter and currently clearly identified.
60 days from the February low. But it gets even better. During the spring and summer of 2010 the Nasdaq-100 fell
to a significant low on the 1st of July. In figure 14 we see that the
count starts from this low, and how the Yearly cycle has been
Figure 13
present. In figure 15 I have zoomed in the latest months and
started the count from the low on the 16th of November. Going
forward it will be very interesting to see how and if this cycle
continuous to vibrate as distinctly as before. Coming up next is
the anniversary of the low from summer of 2010; so now I am
monitoring how this index moves into the time window around
the 1st of July 2011.

Figure 15

Figure 14

More on Gann
I have been studying and working with Gann Analysis
for over a decade and each year and with this, I have gained
additional crucial insight and understanding of time and price.
For those interested in more background, inspiration and
understanding of Gann’s work Bill McLaren’s work, which has
been of great help to me, and can be found on his web site,
Note where AstraZeneca topped in April: At 401 SEK, the www.mclarenreport.net.au.
same level as the top in November the year before. What does Additionally, for any student who wants to go deeper in to
this mean? That it rose 401-266 SEK = 135 SEK in 60 days this Gann’s work, I must recommend David Keller’s edited book
time. Price and time in harmony again! And if we look even Breakthroughs in Technical Analysis – New Thinking from the
closer, the top in March was at 386 SEK which means that the World’s Top Minds. In this book you will find chapter five, by
stock rose 120 SEK in the first 30 days of this uptrend. How Constance Brown and titled Price and Time. Brown makes an
much did it then reverse? From 386 down to 341 = 45 SEK, and insightful presentation of Gann’s work and the man himself.
from here it then rose 60 SEK up to the top at 401 SEK in April.
When Price and time moved in harmony, Gann referred Bibliography
to this as the “squaring” of price and time. That is taking the McLaren W, Gann Made Easy, Gann Theory Publishing, 1986, pp. 37-38.

analysis further and a subject for another article and beyond Keller D, Breakthroughs in Technical Analysis – New Thinking from the
World’s Top Minds, Bloomberg Press, 2007, pp. 83-113.
this short study.
Charts and Data
Trading Session 2

PAGE 26 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

Moving Averages 3.0

by Manfred G. Dürschner

Abstract fluctuations of price movements are replicated in smoother

The well-known Moving Averages (MA), namely the Simple and clearer patterns. Yet smoothing cannot avoid the main
Moving Average (SMA), the Exponential Moving Average (EMA) drawback of SMA: a time lag between the pattern of the price
and the Weighted Moving Average (WMA), are modified in this series and a MA itself. This can clearly be seen if you consider
paper with the help of the Nyquist Criterion. These modified trend reversals. The reversal of a MA lags behind that of the
Moving Averages 3.0 show good smoothing characteristics, price series.
illustrate relevant trends and trend reversals in price series The lags for the different SMA with a cycle period n are
without a time lag as far as calculated. With regard to calculated as follows [1]:
smoothing, trend patterns and time lag bring about a significant
improvement on conventional SMA (Moving Averages 1.0: SMA, SMA: lag = (n – 1)/2,
EMA and WMA). In addition to this, the efficiency of the Moving
Averages 3.0 is demonstrated by applying several tests and a EMA: lag = 1 𝑎 – 1, for a = 2/(n+1), we get the SMAe result as
simple trading system. for the SMA,

Introduction WMA: lag = (n – 1)/3.

In Technical Analysis: SMA are the most widely used It is noticeable that the WMA shows the smallest lag.
indicators. Applied to a price series, the market situation
described as a fluctuating price pattern is then smoothed as Approaches to reduce time lag
the chaotic price fluctuations are smoothed out. Smoothing In 1994, Patrick Mulloy made an innovative approach to
is the most valued advantage of a Moving Average (MA). The reduce the lag [2]. According to the following expression:

Figure 1

US: S&P 500, GD, TEMA, MMAGD

1390 1390

1380
Figure 1 1380

1370 1370

1360 1360

1350 1350

1340 1340

1331,10
1330 MMAW (Price | 50) 1331,10
1330

1320 1320

1310 1310

1300 1300

1290 1290

1280 1280

1270 1270

1260 1260

Tema (Price | 50)

1250 1250

1240 1240

1230 1230

1220 1220

1210

1200
SMA(Price | 50) 1210

1200

1190 1190

1180 1180

1170 1170

1160 1160

1150 1150

1140 1140

1130 1130

1120 1120

1110 1110

1100 1100

1090 1090

1080 1080

1070 1070

1060 1060

1050 1050

1040 1040

1030
Price series: S & P 500 (September 2010 – Mai 2011) 1030

23 30 7 13 20 27 4 11 18 25 1 8 15 22 29 6 13 20 27 3 10 18 24 31 7 14 22 28 7 14 21 28 4 11 18 25 2 9 16 23 30
September '10 Oktober '10 November '10 Dezember '10 Jan 2011 Februar '11 März '11 April '11 Mai '11

IFTA.ORG PAGE 27
 IFTA JOURNAL 2012 EDITION

TEMA = 3*EMA – 3*EMA[EMA] + EMA[EMA(EMA)] Nyquist Criterion

In signal processing theory, the application of a MA to itself
He applied a EMA once and twice to itself and combined the can be seen as a Sampling procedure. The sampled signal is the
results with the original EMA. MA (referred to as MA1) and the sampling signal is the MA as
In 2001, John F. Ehlers made a more general attempt for a MA well (referred to as MA 2). If additional periodic cycles which are
with reduced lag [3]. It runs: not included in the price series are to be avoided sampling must
obey the Nyquist Criterion [1, 4].
MMA = 2 * MA[price | cycle period n] - MA[MA | cycle period n]. With the cycle period as parameter, the usual one in
Technical Analysis, the Nyquist Criterion reads as follows:
Ehlers used a MA (SMA, EMA or WMA) and applied this MA
a second time to itself. This result MA[MA] is subtracted from n1 = λ*n2 , with λ ≥ 2.
the MA multiplied by the factor 2. The MA modified in this way
(notation MMA) is compared with the TEMA in figure 1 (price n1 is the cycle period of the sampled signal to which a sampling
series S&P 500; as MA and EMA is used with a 50 days period; signal with cycle period n2 is applied. n1 must at least be twice as
Ehlers´ MEMA: red line; TEMA: blue line). Ehlers´ simpler large as n2. In Mulloy´s and Ehlers´ approaches (referred to as
relation shows almost the same result as the TEMA as far as the Moving Averages 2.0) both cycle periods are equal.
trend reversals are concerned. Both MAs are comparable as to
their smoothing behavior. In both MAs – MEMA and TEMA – Moving Averages 3.0
there is a clear improvement concerning lag compared to EMA Using the Nyquist Criterion there is a relation by which the
(black line). application of a MA to itself can be described more precisely. In
In the two approaches a MA is applied to a price series and to figure 2 a price series C (black line), one MA (MA1, red line) with
itself. If one considers the price series and the MA in a general lag L1 to the price series and another MA with lag L 2 to MA1 (MA 2,
way as a time-dependent time series, the application of a MA to blue line) are illustrated. Based on the approximation and the
a MA as a sampling procedure and takes findings from the field relations described in figure 2 the following equation holds:
of signal processing, it can be deduced that the application of
a MA to itself (as to MMA and TEMA, see above) is at best only (1) D1/D2 = (C – MA1)/(MA1 – MA 2) = L1/L2
approximately correct. This can be substantially improved by
the Nyquist Criterion. According to the lag formulas in the introduction L1/L2 can be
written as follows:

FigureFigure
2 2

Price series C

MA1 = MA1( price | cycle period p1)

D1
φ1

L1 D2 In a fairly good approximation holds:

φ2 φ1 = φ2 .
L2
Then follows: D1/L1 = D2/L2 ,
with D1 = K – MA1 and D2 = MA1 – MA2 .

MA2 = MA2 ( MA1 | cycle period p2)

PAGE 28 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

α := L1/L2 = (n1 – 1)/(n2 – 1). Test of NMA

In figure 3 a New Weighted Moving Average (NWMA) (a
In this expression denominator 2 for the SMA and EMA as WMA is used for the MA) is compared with Ehlers´ MWMA.
well as denominator 3 for the WMA are missing. α is therefore In both cases a WMA with the cycle periods 21 and 200 days,
valid for all three MAs. Using the Nyquist Criterion one gets for respectively was chosen. In addition, a short period is necessary
α the following result: for the NWMA: 5 days (λ=4.2, 21-day NWMA) and 50 days (λ=4,
200-day NWMA). The German index DAX stands for the price
(2) α = λ* (n1 – 1)/(n1 – λ). series as an example.
From figure 3 you will see:
α put in (1) and C replaced by the approximation term NMA, ƒƒ The NWMA is significantly closer to the price series than the
the notation for the new MA, one gets: MWMA.
ƒƒ The long-term trend of the 200-day NWMA is precise and
NMA = (1 +α) MA1 – α MA 2. close to the price series.
ƒƒ The distinct trend reversals of the 21-day NWMA are
In detail, equation (2) reads as follows: described much more precisely than by the MWMA, and the
(3) NMA[price/ n1, n2] = (1 + α) MA1[price/ n1 ] – α lag of the NWMA against the price series comes to less than
MA 2[MA1/ n2], two days.
ƒƒ In case of 21-day cycle period the NWMA shows a lag which
(4) α = λ* (n1 – 1)/(n1 – λ), with λ ≥ 2. is 3 days shorter than that of the MWMA (compare the two
colored arrows).
(3) and (4) are equations for a group of MAs (notation: Moving
Averages 3.0). They are independent of the choice of an MA. As In summary, it can be concluded that the Moving Averages
the WMA shows the smallest lag (see introduction), it should 3.0 on the basis of the Nyquist Criterion bring about a significant
generally be the first choice for the NMA. improvement compared with the Moving Averages 2.0 and 1.0.
n1 = n2 results in the value 1 for α and λ, respectively. Then Additionally, the efficiency of the Moving Averages 3.0 can be
equation (3) passes into Ehlers´ formula. Thus Ehlers´ formula proven in the result of a trading system with NWMA as basis.
is included in the NMA formula as limiting value. It follows from
a short calculation that the lag for NMA results in a theoretical Trading system based on NWMA
value zero. The trading system consists of one technical indicator, but

Figure 3
DE: DAX, MMAGD, MMAGD-2, ANLMA-2, ANLMA-3

7250

7200
Figure 3 7200
7178,29
7150
7100
7100

7050
7000
7000

6950

6900 NWMA (Price/200, 50) 6900

6850
6800
6800

6750
6700
6700

6650
6600
6600

6550 6500
6500

6450 6400

6400
MMAW (Price/200) 6300
6350

6300
6200
6250

6200
6100
6150

6100
MMAW (Price/21) 6000
6050

6000
5900
5950

5900
5800
5850

5800 NMAW (Price/21,5) 5700

5750

5700 5600
5650 Price series: DAX (April 2010 – February 2011)
6 12 19 26 3 10 17 24 31 7 14 21 28 5 12 19 26 2 9 16 23 30 6 13 20 27 4 11 18 25 1 8 15 22 29 6 13 20 27 3 10 17 24 31 7
April 2010 Mai '10 Juni '10 Juli '10 August '10 September '10 Oktober '10 November '10 Dezember '10 Jan 2011 Februar '11

IFTA.ORG PAGE 29
 IFTA JOURNAL 2012 EDITION

with some significant details: ƒƒ WMA [price series | 89],

ƒƒ SMA [price series | 89].
ƒƒ The indicator is the Aroon-Oscillator (AO), which is defined as
the difference between the Aroon up and Aroon down. The first modification is meant to be a stability test for
ƒƒ The AO is not applied to a price series but to a NWMA applied the NWMA [ price series | 89, 21] system and the further
to the price series: NWMA [ price series | n1, n2 ]. modifications should be compared with Ehlers´ approach and
ƒƒ Cycle periods for the NWMA are n1 = 89 days and n2 = 21 days the standard moving averages WMA and SMA.
(λ = 4.2). The trading system described was tested with 104 selected
ƒƒ Cycle period for the AO is 5 days: AO [NWMA | 5]. shares (Europe, USA and Asia, and 18 different sectors according
ƒƒ An Inverse Fisher Transformation (IFT) is applied to the AO: to DJ Sector Titans) and submitted to a backtesting (Software:
IFT [AO]. Investox, Version 5.9.4). The following data were chosen:
ƒƒ The IFT digitizes the AO without lag.
ƒƒ Settings: Buy IFT > 0, sell IFT < 0. ƒƒ Covered period: January 03, 2000 - January 31, 2011,
ƒƒ Seed capital/share: EUR 1 000.-,
In figure 4 you can recognize the trading system: in the upper, ƒƒ Enter-expenses 0.3 % per trade as well as for exit-expenses
red field the digitized AO, and in the lower part of the chart the and slippage.
price series represented by Heikin-Ashi-Candlesticks and the
Bollinger Bands. The purpose of the Heikin-Ashi-Candlesticks After each trade closed, all capital available was reinvested.
and the Bollinger Bands is to visually monitor trends and The results of average values per share are presented in
volatility. Long-trades are indicated by the green horizontal bars tabular form (see Table 1):
in the lower part of the chart. For examples, three trades are
marked by green (enter) and red (exit) vertical lines. ƒƒ The NWMA trading system shows the highest net profit.
Furthermore, the trading system described was compared ƒƒ Compared with a buy-and-hold-strategy the NWMA trading
with systems which use other MAs instead of the NWMA [ price system described is significantly more profitable over others.
series | 89, 21] (see above). The following modifications were ƒƒ The drawdown numbers of the different trading systems
tested for comparison (no changes due to settings): differ only slightly, with the exception of the SMA-system.
ƒƒ Due to the net profit the NWMA trading system has the best
ƒƒ NWMA [ price series | 100, 25], drawdown.
ƒƒ MWMA [price series | 89], ƒƒ The largest profit together with the highest number of

Figure 4

Figure 4 Price series: DAX (as figure 3)

PAGE 30 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

profitable trades of the NWMA trading system can be explained References

by its quick reaction and the minimal lag of the NWMA. [1] John F. Ehlers: Rocket Science for Traders (John Wiley & Sons, 2001)
[2] Patrick Mulloy: Stocks & Commodities Magazine (February 1994)
[3] John F. Ehlers: Signal Analysis Concepts (Internet, 2001)
Conclusion [4] Wikipedia: Nyquist-Shannon-Criterion
The group of Moving Averages 3.0 is a significant
improvement over conventional Moving Averages 1.0 (SMA, EMA Charts
and WMA) and the well-known approaches to reduce time lags Investox; knöpfel Software GmbH

(Moving Averages 2.0). The NMA is suitable for intraday trading

as well. A system with a Stochastic-RSI-Indicator (cycle periods
3 and 5) applied to a NWMA (cycle periods 8 and 3), time base
15 minutes, shows profitable trades, too (see figure 5 with two
trades as an example).

Figure 5

Table 1

Average values/Share NMAW(89/21) NMAW(100/25) MMAW(89) MAW(89) MAS(89)

Number of Trades 69,3 62,7 33,2 24,8 20,2

Trades/y 7,4 6,7 3,6 2,6 2,2
Net-Profit (EUR) EUR 46 974,57 EUR 32 292,98 EUR 12 317,75 EUR 8 392,35 EUR 4 552,98
Buy/Hold-Profit (EUR) EUR 1 103,09 EUR 1 103,09 EUR 1 103,09 EUR 1 103,09 EUR 1 103,09
Profitable Trades (%) 61,24% 60,68% 60,09% 61,39% 58,77%
Max. Drawdown (%) -8,05% -7,94% -9,43% -7,70% -14,72%

IFTA.ORG PAGE 31
Master of Financial Technical
Analysis (MFTA) Program
IFTA’s Master of Financial Technical Analysis
(MFTA) represents the highest achievement and Examinations
recognition by peers in the Technical Analysis In order to complete the MFTA and receive your
community. Diploma, you must write a research paper of no
less than three thousand, and no more than five
MFTA is open to individuals who have attained thousand, words. Charts, Figures and Tables may
the Certified Financial Technician (CFTe) be presented in addition.
designation or its equivalent, including:
Your paper must meet the following criteria:
• Chartered Member of the Nippon Technical
Analysts Association (CMTA) from the Nippon • It must be original
Technical Analysts Association (NTAA) • It must develop a reasoned and logical
• Diploma in Technical Analysis (Dip.TA) from argument and lead to a sound conclusion,
the Australian Technical Analysts Association supported by the tests, studies and analysis
(AATA) contained in the paper
• Certified ESTA Technical Analyst Program • The subject matter should be of practical
(CETA) from the Egyptian Society of Technical application
Analysts (ESTA) • It should add to the body of knowledge in the
discipline of international technical analysis
MFTA requires an original body of research. It
is intended to be a rigorous demonstration of
professionalism in the global arena of Technical
Timelines & Schedules
Analysis. There are two MFTA sessions per year, with the
following deadlines:
For those IFTA Colleagues who do not have
the formal qualifications outlined above, but Session 1
who have other certification and/or many years “Alternative Path” application deadline
experience working as a technical analyst, February 28
the Accreditation Committee has developed Application, outline and fees deadline
an “alternate path” by which candidates with May 2
substantial academic or practical work in Paper submission deadline
technical analysis, can bypass the requirements October 15
for the CFTe, and prequalify for the MFTA.
Session 2
There are three categories of applicant for the “Alternative Path” application deadline
alternate path. It is open to individuals who have: July 31
• A certification such as Certified Market Application, outline and fees deadline
Technician (CMT), Society of Technical Analysts October 2
(STA) Diploma, PLUS three years experience Paper submission deadline
as a technical analyst; or March 15 (of the following year)
• A financial certification such as Certified
Financial Analyst (CFA), Certified Public
Accountant (CPA), Masters of Business
To Register
Administration (MBA) PLUS five years Please visit our website at http://www.ifta.org/
experience as a technical analyst; or certifications/master-of-financial-technical-
• Have a minimum of eight years experience as analysis-mfta-program/ for further details and to
a technical analyst. register.

Candidates in these circumstances may apply

for the “alternate path”. If approved, they may Cost
register for the MFTA and send in their research $900 US (IFTA Member Colleagues);
proposals. $1,100 US (Non-Members)
 IFTA JOURNAL 2012 EDITION

Applying Trading Strategies to Price Channel Breakout

Trading—Statistical Significance of Channel Breakout Variation
by Robin Boldt, MFTA

Abstract on closing prices.1 It might seem to be just a minor detail and

Chart patterns are an important aid in technical analysis, overlooked easily, but from a trading perspective it is a very
but only when they give high probability outcomes. Hence the different approach to wait for closing prices as a breakout
key feature of any chart pattern should be reliability. The price confirmation. Odds could favor a valid breakout if the closing
channel by Richard Donchian can be classified as one pattern price is above the previous n-period high (low), but in reality
that is said to be reliable. It is a common breakout system as new once the price moves away from the pattern, the trader might
highs and lows are a readily available piece of information for be late buying into the breakout as the risk reward ratio starts
every financial market participant, especially technical analysts to narrow, as compared to buying directly into the breakout to
and traders. The purpose of this paper is to test the profitability benefit from the forming trend. In addition—and at the risk of
of price channel variations and to investigate what different a pedantic view the practicability of the buy (sell) on the close is
outcomes in profitability are resulting from. The key question a critical consideration as the trader does not know in advance
is: When are breakouts in the Donchian system to be bought if the closing price will really describe a new n-period high or
and which additional parameters influence its profitability? low. This problem emerges especially when trading European
Research has put a lot of effort in finding the most profitable stocks as their closing price is determined by a closing auction,
price pattern, however little research has been published on the hence their closing price can differ materially from any pre-
right timing within trend following strategies applied to stocks. closing indication and still may not make it to a new high or low.
While I cannot rule out the possibility that technical analysis For this reason, buy and sell signals on closing prices within
complements other profitable market timing techniques or this paper are based on buy (sell) on open of next bar if the
trading rules, I decided to put four hypotheses to the test by previous bar describes a completed n-period high (low). Finally, I
running them against a comprehensive database of German examined short selling signals based on price channel breakouts
stocks. Realistic transaction cost estimates were applied. This (breakdowns) in equal measure using the lower price channel
paper illustrates results testing different parameters on price bands as a stop level with momentum strategies.
channel breakouts. The empirical results strongly suggest Despite positive evidence in literature about profitability,
that easy to use trading parameters and rules can increase the purpose of this paper is to test if other easy-to-use-trading
profitability. variations and rules actually do add value, which will hopefully
lead to practical methods to identify and trade successful
Part one: Introduction breakouts.
The following chart (Figure 1) is a perfect example of a
1.1 Presenting the Hypotheses Donchian price channel breakout where Novartis breaks
Price pattern breakouts are one of the most common forms above its 10-day price channel (breakout #1) on an intraperiod
of breakouts as they are based on pure price action. Donchian (intraday) and closing price. Breakout #2 is slightly different as
breakout signals are generated whenever the market price is the stock recovered shortly after the breakout and completed
equal to or greater than the highest high or the lowest low of a strong intraday reversal on its closing. Let us assume here
the past n periods. There are many types of breakout patterns, that the trader bought into the stock on the back of breakout
such as triangles, channels, wedges, pennants and flags. This #1 as per the closing or opening of next bar. When should he
work, however, will focus on only the Donchian breakout close his position? Apparently as soon as the next breakout
signals. Numerous empirical studies have proven the general to the downside occurs negating the initial buy signal. His
profitability of price channel breakout strategies but there is performance will differ substantially depending on what his
a necessity to further broaden the research spectrum on this exit strategy, e.g. on intra-period price channel breakouts
topic. The focus of research has been skewed to the futures (breakout #2) or price channel breakouts with a closing price
market including forex and commodities, and one geographic confirmation (breakout#3). The same applies for his entry
region: North America. In addition, performance assessments strategy: Apparently, buying directly into the breakout as it
were concerned to find the most profitable periods—understood happens versus not buying on the closing or opening of next bar
as days used to plot the upper and lower bands—instead of at provides a much better entry as a tighter stop can be placed
setting different timing parameters in direct contrast, i.e. price as opposed to a delayed entry—where the breakout is only
channel breakout strategies based on closing prices, versus bought after a closing price, outside the price channel. Executing
price channel breakout signals based on intra-period lows/ a trade on breakouts are a zone of conflict and timing.
highs in comparison to various risk and performance metrics. Before reviewing the analysis, I want to emphasize what this
For example, Chande’s price channel breakout test is based work is and is not: This work does not claim to cut the Gordian

IFTA.ORG PAGE 33
 IFTA JOURNAL 2012 EDITION

Figure 1: Novartis (daily candle chart, 10-day Donchian price channel) Source: Reuters (2011)

knot to find the most profitable price channel, but intends to will chart the level of the highest high in the last 10 days above
emphasize and demonstrate that a successful trading strategy the price line, and will chart the level of the lowest low in the
is not built upon a single, but many factors; and that by changing last 10 days below the price line.2 Thus, price channels use
the model’s parameters its profitability can be increased maximum and minimum price values and not moving averages
significantly.   or standard deviations as boundaries. If the most recent price is
To give a new light to the issue, this paper will test the a new n-period high (low), it will be charted outside of the price
effectiveness of the following four hypotheses, which should channel and typically generates a buy (sell) signal at the point
lead to practical and working answers on how to identify of the breakout (a price channel breakout). The price channel is
successful breakouts: a trend following breakout system: Buy when price moves above
the channel, sell when the price moves below the channel.
1. Price channel breakout buy and sell signals based on closing
prices of completed trading periods are (significantly) less 1.3 Related Work and Similar Concepts
profitable than buying or selling intraperiod highs or lows “Buy strong action, sell weak one.” 3 - The use of price
directly. channels on charts to generate buy and sell signals is anything
2. Price channel breakout sell signals are as profitable as buy but new. Richard Donchian developed the Donchian 4 Week
signals. Breakout Channel strategy in the 1960s whereby signals are
3. The smaller the number of periods charting a n-period generated whenever the price exceeds the highs of the four
high or low around the price line, the more profitable price preceding calendar weeks. Like all other trend followers,
channel breakout signals are. Donchian emphasized the importance of the price: He does
4. Price channel breakouts that occur on low volume cannot be not predict price movements; he just follows them as “trends
considered unreliable.  persist” .4 John J. Murphy and other technical analysts made
adjustments to the four week rule, for example by shortening
1.2 Price Channel Breakouts and lengthening the time periods for sensitivity.5 The chart
The price channel, or Donchian channel, is a volatility below is a good example of a price channel breakout on a longer
indicator, calculating the recent price range using the most time horizon in which we see the S&P 500 Index breaking
recent high and low to mark the high and low bands (creating a above its four month price channel in early 1995 to give a buy
channel containing the prices). The Donchian channel therefore signal, which remained intact for years. The basic idea of the
has a similar chart appearance to other volatility indicators trading system is continuous in nature, which means that the
such as Bollinger Bands. For example, a 10-day price channel trader always has a position, either long or short, rendering it

PAGE 34 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

vulnerable to whipsaws during trendless markets.6 Richard 36% winning trades on average and a win/loss ratio of 2.01.
Dennis was one who attempted to break this continuum, as Additional empirical studies proved the profitability of the price
his famous turtle trading rules are based on buy (sell) signals channel breakout, such as Lukac et. al (1990).
generated on breakouts above (below) the high (low) of the last
20 bars while using an exit depending on the highest high (for Part Two: Test-Design
shorts) and lowest low (for longs) of the last ten bars.7 Tushar
Chande tested how the originally continuous price channel 2.1 Overview
breakout system performs if modified to a noncontinuous To recall, the following hypotheses are to be tested:
system by examining the effect of adding a trailing stop (exit
on the highest high or lowest low of the last five days) while 1. Price channel breakout buy and sell signals based on closing
also modifying the entry rule: “If today’s close is higher (lower) prices of completed trading periods are (significantly) less
than the highest high (low) of the last 20 days, then buy (sell) profitable than buying or selling intraperiod highs or lows
on the close and exit the long trade at the lowest low of the last directly.
five days on a stop”.8 This modification naturally increases the 2. Price channel breakout sell signals are as profitable as buy
quantity of trades and as a consequence results into higher signals.
trading costs, making the strategy vulnerable to whipsaws. 3. The smaller the number of periods charting a n-period
Overall, his tests showed profitable trading results with high or low around the price line, the more profitable price
channel breakout signals are.
4. Price channel breakouts that occur on low volume cannot be
Figure 2: S&P 500 (monthly bar chart) Source: John J. Murphy
(1999), “Technical Analysis of the Financial Markets”, p. 220 considered unreliable. 

In a first run, a testing of hypotheses 1 and 2 will be

conducted (Test #1) where the same breakout system is run on
two different entry signals (all other things being equal) using a
ten day period to calculate the price line. The testing will help to
emphasize the importance of the right timing strategy in place
as many trading strategies are (wrongfully) deemed to be not
profitable without looking at those two parameters. In addition,
the performance of long and short signals is compared.
Hypotheses 3 will be tested through Test #2 using the most
profitable trading setup emerging from Test #1.
The purpose of Test #3 is to assess the importance of volume
during price channel breakouts, by filtering out breakouts that
occur on low volume. The question here is: Are price channel
breakouts that occur on low volume significantly less profitable?

Table 1: Test #1 – Paired t-test results

Setup 1 Setup 2
All Long Short All Long Short
Mean Net Profit per Stock 13,865.26 6,029.32 7,835.94 4,179.72 1,462.84 2,716.88
Gross Profit 26,718.92 13,394.28 13,324.64 12,590.22 5,757,35 6,832.87
Profit Factor 2.49 2.21 3.33 2.19 2.23 4.53
Total Number of Trades 202 105.50 96.77 81.70 43.43 38.27
Percent Profitable 33% 33% 34% 36% 33% 40%
Average Trade Net Profit 66.45 55.79 78.05 45.68 30.96 68.17
Average Winning Trade 381.50 373.99 389.52 412.13 390.67 430.03
Average Losing Trade -92.10 -100.58 -82.69 -153.64 -141.62 -166.05
Ratio Avg.Win./Avg.Los. 4.74 4.26 6.14 3.50 3.98 5.56
Largest Winning Trade 1,025.69 810.08 934.77 1,017.98 676.03 937.95
Largest Losing Trade -2,128.72 -2,045.11 -1,612.51 -2,006.86 -1,787.95 -1,702.40
Total Commission 10,562.28 5,865,82 4,696.46 3,885.28 2,199.58 1,685.69
Max. Drawdown -3,343.72 -2,717.74 -2.164.90 -3,609,01 -2,692.74 -2,415.83
Max. Intraday Drawdown -3,920 -3.237.37 -2,702.32 -4,241.17 -3,431.82 -2,945.00
Total Positions 209.1 6,029,32 7,835.94 84.80 45.03 39.77
Position Changes 418.2 13,394.28 13,324.64 169.60 90.07 79.53

IFTA.ORG PAGE 35
 IFTA JOURNAL 2012 EDITION

Table 2: Test #1 – Paired t-test comparing the mean net profit

per stock price for sells respectively. Of course, the bid and ask price is an
estimation of a realistic execution price.
Setup 1 Setup 2
Mean performance 13,865.26 4,179.72 2.4 Money Management
Mean 1 – Mean 2 9685.54 Each trading portfolio per stock is set up with an initial
N 30 30 capital of €100,000. The size of an order is not fixed but depends
on the risk attached to each buy and sell signal. The risk per
Standard deviation 7,809.6889 5,235.5954
trade is defined by the difference between the entry and the
Standard error of mean 1425.8476 955.8846 stop level. The system does not allow a risk of more than 2% per
df 29 trade of the current capital.
p-value less than 0.0001
t 7.7932 2.5 Supplementary
Statistically significant? Yes Pyramid trading, understood as of using profit generated
from an existing position to acquire additional positions, is not
incorporated in the empirical testing. Hence, there is no adding
Table 3: Test #1 – Paired t-test comparing the average net to an existing position.
profit per trade
Setup 1 Setup 2 Part Three: Backtesting and Empirical
Mean performance 66.45 45.68 Results
Mean 1 – Mean 2  20.76
N 30 30 3.1 Test #1
Test #1 aims to test the profitability of price channel
Standard deviation 37.0400 58.3895
breakout trading using two different set-ups:
Standard error of mean 6.7626 10.6604 Setup 1: Entry: Buy (sell short) at market if the stock hits a
df 29 fresh 10-period high (low).
p-value 0.0481 Setup 2: Entry: Buy (sell short) on next bar when stock’s close
t 2.0639 is above its n-period high (low).
Statistically significant? Yes All other parameters are held equal: exit long (short) position
on stop when stock breaks below the trailing price level of the
lower (upper) price channel or when the break-even level is
2.2 Statistical Significance of Channel Breakout touched. The stop loss is set to breakeven once the open profit
Variations exceeds 2€ so that there no longer risk attached to the trade.
The key question of the upcoming price channel breakout The exit is defined as follows: Take profits when 20% of the
variations is whether the differences in the profitability are initial risk is earned.
statistically significant or not. In this paper, profitability is Setup 1 yields on average—as seen in table 2 – 13,865€ per
understood as mean net profit per trading strategy and average stock over the three year period, which is significantly higher
net profit per trade. Broadly speaking, a test of significance is a than in setup 2. Further testing elaborates that this increased
procedure by which sample results are used to verify the truth profitability is a result of a higher trade frequency (202 vs. 81
or falsity of a null hypothesis (H0).9 The decision to accept or trades on average) and also a significantly higher average net
reject H0 is made on the basis of the value of the test statistic profit per trade, which is an important finding.
obtained from the data at hand. I will use a paired t-test in this
paper as the measurements are taken from the same stock 3.1.1 Supplementary: Test #1.2—Long vs. Short Trades
universe. By using the paired sample t-test, I can statistically Additional test results, presented in table 4 and 5, clearly
conclude whether adjustments to certain price channel trading show significant profitable short sells signals based on price
parameters have improved the profitability or not. channel breakdowns, are and that the signals are not to be
neglected:
2.3 Data and Transaction Costs
The scope of this test includes stocks that are DAX members 3.2 Test #2
as per September 2010. This amounts to 30 securities spanning Hypothesis 2 will be tested through Test #2, using the most
three years as the performance will be tested on historical data profitable trading strategy emerging from Test #1 which is
from 1/1/2004 through 12/31/2009. The reason I chose these setup 1. For the purposes of this test, the chosen entry method
stocks and that timeframe is twofold: First, to avoid ill-liquid was a 5-, 10- and 20-period high and low on a daily chart (all
stocks where realistic historical results could not be generated. other parameters hold equal). I intentionally did not include any
Second, to assure a test window large enough to generate timeframe optimization to avoid overfitting the data. The aim
statistically sound results, including a broad data sample and is to observe a general change in profitability if the number of
conditions. Realistic transaction cost estimates were applied in periods that will chart a n-period high or low around the price
the amount of 0.10% per transaction side. To avoid unrealistic line are changed.
fills, the execution price is conducted at ask price for buys, or bid Setup 3: Buy (sell short) at market if the stock hits a fresh

PAGE 36 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

Table 4: Test #1 – Paired t-test comparing the mean net profit per stock
Setup 1 Shorts Setup 1 Longs Setup 2 Shorts Setup 2 Long
Mean performance 7.835,94 6.029,32 2.716,88 1.462,84
Mean 1 – Mean 2 1806.6150 1254.0317
N 30 30 30 30
Standard deviation 4424.5568 4453.8320 3496.6762 3145.5346
Standard error of mean 807.8099 813.1547 638.4028 574.2934
df 29 29
p-value 0.0262 0.1048
t 2.3431 1.6742
Statistically significant? Yes No

Table 5: Test #1 – Paired t-test comparing the average net profit per trade
Setup 1 Shorts Setup 1 Longs Setup 2 Shorts Setup 2 Long
Mean performance 55,79 78,05 30,96 68,17
Mean 1 – Mean 2 -22.2557 -37.216
N 30 30 30 30
Standard deviation 40.3487 42.7422 74.0995 94.2504
Standard error of mean 7.3666 7.8036 13.5287 17.2077
df 29 29
p-value 0.0035 0.0644
t 3.1739 1.9226
Statistically significant? Yes No

Table 6: Test #2 – Paired t-test results Table 8 illustrates that the mean net profit per stock does not
Setup 3 Setup 4 Setup 5 result from a statistically significant higher net profitability per
trade, but from an increased trading frequency due to a smaller
number of periods charting an n-period high and low around the
Mean Net Profit per Stock 21,622.76 13,865.26 8,977.34 price line, which trigger more buy and sell signals (table 7). The
Gross Profit 47,876.75 26,718.92 15,245.41 trade frequency (number of trades: 304 vs. 202 vs. 132) is the
Profit Factor 2.10 2.49 3.14 key performance driver and not the average net profitability per
trade (not statistically different).
Total Number of Trades 304.80 202 132.03
Percent Profitable 44% 33% 26% 3.3 Test #3
Average Trade Net Profit 67.46 66.45 61.27 The purpose of Test #3 is to assess the importance of volume
Average Winning Trade 341.46 381.50 406.79 during price channel breakouts, by filtering breakouts that
Average Losing Trade -147.92 -92.10 -62.33 occur on low volume. The question here is: Should price channel
breakouts that occur on low volume be avoided?
Ratio Avg.Win./Avg.Los. 2.55 4.74 8.25 Setup 6: Buy (sell short) at market if the stock hits a fresh
Largest Winning Trade 1,158.80 1,025.69 874.03 5-period high (low).
Largest Losing Trade -2,599.54 -2,128.72 -1,667.40 Setup 6v: Buy (sell short) at market if the stock hits a fresh
Total Commission 23,524.39 10,562.28 4,994.25 5-period high (low) but only when volume exceeds its five day
volume exponential moving average (EMA).
Max. Drawdown -5,715.01 -3,343.72 -2,343.05
Setup 7 and 8 are structured as in Test #2. All things hold
Max. Intraday Drawdown -6,227.29 -3,920 -2,892.24 equal for Setup7v and Setup8v except for a volume threshold as
Total Positions 315.47 209.10 136.90 in Setup 1v, using a 10 and 20day volume EMA respectively.
Position Changes 630.93 418.20 273.80 The results presented in table 10 and 11 provide much insight:
I have always been a fan of breakouts on relatively “decent”
or high volume but the test results tell a different story: Price
5-period high (low). channel breakouts that occur on low volume should not be
Setup 4: Buy (sell short) at market if the stock hits a fresh avoided. Trading price channel breakouts only when they occur
10-period high (low). on relatively higher volume will have a negative impact on the
Setup 5: Buy (sell short) at market if the stock hits a fresh total profit and loss of this signal, because breakouts that occur
20-period high (low). on average volume are also reliable(see table 9).

IFTA.ORG PAGE 37
 IFTA JOURNAL 2012 EDITION

Table 7: Test #2 – Paired t-test comparing the mean net profit per stock
Setup 3 vs. Setup 4 Setup 4 vs. Setup 5 Setup 3 vs. Setup 5
Mean performance 21,622.76 13,865.26 13,865.26 8,977.34 21,622.76 8,977.34
Mean 1 – Mean 2 7,757.50 4,887.92 12,645.43
N 30 30 30 30 30 30
Standard deviation 17,273.23 7,809.68 7,809.68 9,893.87 17,273.23 9,893.87
Standard error of mean 3,153.6461 1,425.8476 1,425.8476 1,806.3656 3,153.6461 1,806.3656
df 29 29 29
p-value < 0.0001 0.0057 < 0.0001
t 3.7204 2.9854 4.8476
Statistically significant? Yes Yes Yes

Table 8 Test #2 – Paired t-test comparing the average net profit per trade
Setup 3 vs. Setup 4 Setup 4 vs. Setup 5 Setup 3 vs. Setup 5
Mean performance 67.46 66.45 66.45 61.27 67.46 61.27
Mean 1 – Mean 2 1.0077 5.1787 6.1863
N 30 30 30 30 30 30
Standard deviation 53.4428 37.0400 37.0400 44.3606 53.4428 44.3606
Standard error of mean 9.7573 6.7626 6.7626 8.0991 9.7573 8.0991
df 29 29 29 29 29 29
p-value 0.8556 0.4179 0.4528
t 0.1836 0.8218 0.7609
Statistically significant No No No

Table 9: Test #3 –Test results

Setup 6 Setup 6v Setup 7 Setup 7v Setup 8 Setup 8v
Mean Net Profit per Stock 21,622.76 9,346.10 13,865.26 7,239.90 8,977.34 3,683.61
Gross Profit 4,7876.75 21,304.98 26,718.92 12,625.42 15,245.41 6,968.20
Profit Factor 2.10 2.70 2.49 3.15 3.14 3.46
Total Number of Trades 304.80 146.90 202 98.70 132.03 65.83
Percent Profitable 44% 43% 33% 34% 26% 27%
Average Trade Net Profit 67.46 63.36 66.45 71.48 61.27 55.20
Average Winning Trade 341.46 329.44 381.50 367.24 406.79 389.99
Average Losing Trade -147.92 -134.82 -92.10 -79.27 -62.33 -64.91
Ratio Avg.Win./Avg.Los. 2.55 3.33 4.74 5.89 8.25 8.80
Largest Winning Trade 1,158.80 883.27 1,025.69 930.83 874.03 778.66
Largest Losing Trade -2,599.54 -1,975.34 -2,128.72 -1,565.67 -1,667.40 -1,238.74
Total Commission 2,3524.39 10,570.54 10,562.28 4,927.73 4,994.25 2,271.81
Max. Drawdown -5,715.01 -3,733.55 -3,343.72 -2,197.05 -2,343.05 -1,694.95
Max. Intraday Drawdown -6,227.29 -4,247.58 -3,920 -2,676.09 -2,892.24 -2,221.90
Total Positions 315.47 151.47 209.10 102.07 136.90 68.07
Position Changes 630.93 302.93 418.20 204.13 273.80 136.13

Discussion and Conclusion In general, price channel breakouts are profitable, however
Price channel breakouts, as well as breakdowns, describe Test  #1 confirmed hypothesis 1 whereby price channel breakout
zones of conflict or “areas of doubt”.10 Quite often price buy and sell signals based on closing prices of completed trading
breakouts turn out to be misleading moves known as whipsaw periods are significantly less profitable than directly buying or
breakouts. Therefore it is crucial for a trader to assess the selling intraperiod highs or lows. Setup 1 (buy at market if the
probabilities of a valid buy or sell signal whenever planning stock hits a fresh 10-period high) offered a higher average
on opening a new position based on a price channel breakout. profit factor (2.49 vs. 2.19), a significant higher average trade

PAGE 38 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

Table 10: Test #3 – Paired t-test comparing the mean net profit per stock
Setup 6 vs. Setup 6v Setup 7 vs. Setup 7v Setup 8 vs. Setup 8v
Mean performance 21,622.76 9,346.10 13,865.26 7,239.90 8,977.34 3,683.61
Mean 1 – Mean 2 12,276.66 6,625.35 5,293.72
N 30 30 30 30 30 30
Standard deviation 17,273.23 7,938.79 7,809.68 4,047.92 9,893.87 3,181.97
Standard error of mean 3,153.6461 1,449.4198 1,425.84 739.04 1,806.36 580.9473
df 29 29 29
p-value < 0.0001 < 0.0001 0.0052
t 5.8185 6.1387  3.0237
Statistically significant Yes Yes Yes

Table 11: Test #3 – Paired t-test comparing the average net profit per trade
Setup 6 vs. Setup 6v Setup 7 vs. Setup 7v Setup 8 vs. Setup 8v
Mean performance 67.4570 63.3550 66.4493 71.4840 61.2707 55.2047
Mean 1 – Mean 2 4.1020 -5.0347 6.0660
N 30 30 30 30 30 30
Standard deviation 53.4428 56.3418 37.0400 40.6389 44.3606 49.0845
Standard error of mean 9.7573 10.2866 6.7626 7.4196 8.0991 8.9616
df 29 29 29
p-value 0.4532 0.3702 0.4434
t 0.7602 0.9102 0.7770
Statistically significant No No No

net profit (66.45€ vs. 49.01€), a better ratio of average win-loss tests within this paper.
ratio (4.74 vs. 3.50) and a statistically significant mean net Test #2 indicates that the smaller the number of periods
profit per stock (13,865.26€ vs. 4,179.72€) compared to setup charting a n-period high or low around the price line, the more
2. Setup 1 presented an even more compelling risk profile with profitable price channel breakout signals are. Profitability,
a lower maximum drawdown (-3,343.72€ vs. -3.609.01€) and a however, is not the only key metric when assessing a trading
smaller average losing trade (-92.10€ vs. -153.64€). Setup 1 and strategy. Setup 5, using a 20-period, offers a superior profit
2 are typical trend following setups with 33% and 36% percent factor (3.14 vs. 2.10 and 2.49), a smaller average losing trade
profitable trades on average. The empirical findings suggest (-62.33€ vs. -147.92€ and -92.10€) and a lower maximum
that there is a payoff to participate in breakouts (sell into drawdown (-2892.24€ vs. -5715.01€ and 3343.72€) as well as a
breakdowns) at an early stage, if combined with tight stops. smaller trading frequency, which explains the lower mean net
Test #1.2 demonstrated that short sells based on price profit per stock over the observed trading period. The higher
channel breakouts (breakdowns) are, interestingly, on average trading frequency in setup 3 may be attractive to traders if their
more profitable than the buy signals, and to some extent trading costs are low. Again, one of the key questions you must
statistically significant more profitable. A reason for that could ask when you see system variations is whether the differences
be attributed to the fact that breakdowns trigger stop-loss fit your trading style. Statistically significant profitability of
orders which then quickly generate selling pressure, whereas channel breakout variations cannot be regarded in isolation.
decisive moves to the upside most often require actively entered Test #3 successfully overcomes the conventional wisdom
buy orders in reaction to the move, compared to automatically that breakouts that occur on low volume can be considered
triggered stop orders. This evidence supports the observation unreliable. With respect to percentage profitable trades or
that many market participants do not feel comfortable to buy average winning trades, it was not at all statistically significant
into a new high by chasing the market as they can hardly assess whether the breakout or breakdown occurred on above average
the probabilities of price continuation, whereby stop losses are volume or not. Breakouts on above average volume show
automatically entered in loss aversion. the same hit ratio but a lower mean net profit, as the volume
Overall, the high frequency of trades suggests that trading threshold halves the number of total trades—resulting in lower
costs are a significant factor in overall performance. It is profitability. In short: On average, a 50% lower trade frequency
noteworthy, that most of the high frequency trading accounts yields a lower mean net profit (ceteris paribus) while both setups
in Europe do not pay more than 0.02% commission (for risk free offer the same mean net profit per trade.
trades) in contrast to the 0.10% commission incorporated in the To summarize, the empirical results strongly suggest

IFTA.ORG PAGE 39
 IFTA JOURNAL 2012 EDITION

that trend following on stocks offer a positive mathematical References

i. Tushar S. Chande (2001) “Beyond Technical Analysis: How to Develop and
expectancy (in line with previous research) and that correctly Implement a Winning Trading System”, Wiley , 2nd edition, p. 203, ff.
timed entries and exits—although often underestimated— ii. Richard L. Weissman (2004) “Mechanical Trading Systems: Pairing
significantly influence the profitability of a price channel Trader Psychology with Technical Analysis”, Wiley, p. 30
iii. Barbara Saslaw Dixon (1978) “Donchian`s 20 guides to trading
breakout. commodities” Commodity Research, p.1
Future research in that regard is requested to expand testing iv. Martin Pring (2004) “Pring on Price Patterns: The Definitive Guide to
Price Pattern Analysis and Intrepretation” McGraw-Hill; 1 edition, p. 88
on other stocks, timeframes and entry and exit rules. v. John J. Murphy (1999) “Technical Analysis of the Financial Markets: A
Comprehensive Guide to Trading Methods and Applications” New York
Bibliography Institute of Finance, p. 216
vi. Brent Penfold (2010) “The Universal Principles of Successful Trading:
1. Chande, Tushar S. (2001) “Beyond Technical Analysis: How to Develop
Essential Knowledge for All Traders in All Markets” Wiley, p. 198
and Implement a Winning Trading System”, Wiley , 2nd edition
vii. Tushar S. Chande (2001) “Beyond Technical Analysis: How to Develop
2. Covel, Michael (2004) “Trend Following: How Great Traders Make
and Implement a Winning Trading System”, Wiley , 2nd edition, p. 203
Millions in Up or Down Markets”, Financial Times Prentice Hall Books
viii. Damodar Gujarati (2004) “Basic Econometrics“ Fourth Edition, The
3. Dixon, Barbara Saslaw (1978) “Donchian`s 20 guides to trading McGraw−Hill, p. 134
commodities” Commodity Research ix. Robert D. Edwards and John Magee, Technical Analysis of Stock
4. Gujarati, Damodar (2004) “Basic Econometrics“ Fourth Edition, The Trends, Amacon, 1948, Pg. 114
McGraw−Hill
5. Kestner, Lars (2003) “Quantitative Trading Strategies: Harnessing
the Power of Quantitative Techniques to Create a Winning Trading
Software and Data
Tradsesignal Standard Edition, Version 5.6.5.0, Copyright 1996-2010
Program” McGraw-Hill tradesignal GmbH
6. Lukac, L.P., B.W. Brorsen, and S.H. Irwin. “A Test of Futures Market MS Office XP
Disequilibrium Using Twelve Different Technical Trading Systems.” Data provided by teletrader and Reuters
Applied Economics, 20(1988):623-639. All programming used to create the trading strategies in the paper was
7. Murphy, John J. (1999) “Technical Analysis of the Financial Markets: A written in the equilla language
Comprehensive Guide to Trading Methods and Applications“ New York
Institute of Finance
8. Pardo, Robert (1992) “Design, Testing, and Optimization of Trading
Appendix (Endnotes)
Systems” John Wiley & Sons Inc 1. Chande (2001) p. 203, ff.
2. Weissman (2004) p. 30
9. Penfold, Brent (2010) “The Universal Principles of Successful Trading:
3. Richard Donchian in 1934, quoted by Dixon (1978), p.1
Essential Knowledge for All Traders in All Markets” Wiley 4. Covel (2004)
10. Pring, Martin (2004) “Pring on Price Patterns: The Definitive Guide to 5. Murphy (1999), p. 216
Price Pattern Analysis and Intrepretation” McGraw-Hill; 1 edition 6. Murphy (1999), p. 216
11. Weissman, Richard L. (2004) “Mechanical Trading Systems: Pairing 7. Penfold (2010), p. 198
Trader Psychology with Technical Analysis”, Wiley 8. Chande (2001) p. 203
9. Gujarati (2004) p. 134

Educated Investor
Plaza Level, 500 Collins Street, Melbourne
urne 3000 Australia
tedinvestor.com.auu
+61 (0)3 9620 0885 [email protected]
www.educatedinvestor.com.au

Australia’s • Technical Analys

Analysis
yssiss

only specialist • Derivatives

investment
• Getting Started
• Superannuation

bookshop • Property
• Business
• Fundamental Analysiss
Established in 1998 to meet the
needs of the investor and trader, • Foreign Exchange

we specialise in books on all asset • Trading

classes, as well as software. • Software

We are pleased to offer a 10% discount on books to all IFTA members, please contact us for details.

PAGE 40 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

Trading the T-Grid

by John Gajewski

Abstract The t-grid Concept and Construction

The t-grid is a series of stacked trend channels with the base The t-grid is a series of stacked trend channels with the
channel defined by the first swing of the trend and determines base channel defined by the first swing of the trend. As the
price vibration, support and resistance ranges, allowing the trend unfolds and price moves beyond the base channel,
trader to create a trade strategy around the t-gric concept. The another channel is added on top of the existing channel. Each
t-grid assists in determinig if price is relatively high or low by new channel line acts as an “attractor” and price advances
breaking up the major trend into smaller defined ranges. With towards it. It will also act as a resistance point. While the trend
application of the t-grid, the trader is presented with effective continues, prices will eventually move through this channel and
and well-defined trading opportunities. a new channel will be stacked on the existing channels.
In this study we step through the process of creating a t-grid Volatility tends to increase as the trend unfolds, so not
from its origin and progress it through its trend development. all new channels will be of equal importance. Rather than
a stacking of channels it is better to view the process as a
Introduction doubling of the channel width. In Figure 1 the darker, solid lines
According to the old adage it is easy to make money trading represent the doubling of the channel. The original channel
the markets—buy low and sell high. But the question when width maintains some value and is represented by a broken line.
is the market low and when is it high is easily answered. It is Just as price is defined by the original swing, so is time. The
relatively easy to define the market as low when it is rising off a vertical lines in Figure 1 are spaced at the time length between
well-defined base, but further into the trend the question if the the first two lows. This time length—let’s call it a vibration—
market is high or low is problematic at best. becomes the basic unit of time for the trend and significant
The trend grid, or t-grid, is a concept providing useful turning points will tend to line up with these vibrations.
insight to answer this question. At the heart of the idea is the This combination of time and price structured by the initial
understanding that the initial swing of a major trend provides swing of the trend provides a powerful method to break up the
the building block for the entire trend. The rules of the t-grid evolving trend and identify key levels and events on which to
provides a systematic way of identifying key levels that define take action. Simple bar analysis will identify price and time
shorter trading ranges, periods in which it is possible to label the levels to open and close trades.
market as low or high. Simple bar analysis can then be used to In this study we will step through the process of setting
identify entry and exit times. up a t-grid from its origin and progress it through its trend
This method favours stocks and commodities that trend development. We will first focus just on the price component.
well and market-wide instruments like the stock indices and
currencies. Large cap stocks also fit the t-grid well, but smaller Drawing the first channel of the t-grid
cap stocks do not as they do not exhibit stable long-term trends
relative to large caps. Commodities produce good t-grids Figure 2: Drawing a channel for the first time
on long-term charts but not so well on shorter time frames.
Whenever an instrument enters a prolonged trend, it is likely to
follow the t-grid structure.
A good example of the t-grid is the Australian Dollar over the
past 12 years.

Figure 1: AUD/USD 2000-2011

A channel can be first drawn with the base trendline

between a low (high) and its retest and the guideline drawn
through the intervening high (low). This simple definition

IFTA.ORG PAGE 41
 IFTA JOURNAL 2012 EDITION

will result in a channel being drawn when there are no firm Figure 3: retest of a previous channel resistance.
signs of a reversal of the previous trend and the action is
considered a pullback rather than the start of a new trend. In
such circumstances it is better to wait for a stronger sign of a
reversal before starting the t-grid. When the t-grid is drawn
it is done so from the origin of the trend. Figure 2 shows the
Australian Dollar starting its t-grid in this fashion.
The low point was made in April 2001 and retested in
September. The intervening high of August was first breached
in April 2002 providing the first opportunity to draw the
channel. However, it was not until June 2002 that the last
swing high of the downward trend was breached, giving us a
firmer indication that a trend reversal may be developing. This
was the time to draw the t-grid.
The t-grid merely indicates key support and resistance
levels for the new uptrend. These are points at which we can Classic t-grid action
rate the market as relatively high and low. In Figure 2 the The next swing shown in Figure 4 demonstrates the classic
Australian Dollar has moved into the higher channel so the within the t-grid.
first lower channel line is support—or the market at a low—and
the higher channel line is resistance—or the market at a high. Figure 4: the rally progresses.
How meaningful this is depends on the width of the immediate
channel in relation to average daily range—the larger the ratio,
the greater the potential opportunity to profit from a trade
within the immediate channel.
We must remember, and within this concept, is that channel
lines act as key psychological points in the market just as
swing highs and swing lows. The channel lines relate price to
time—key points are not just determined by price, but also
have a time dimension. These lines will gain in importance
each time price changes its behaviour as it approaches these
levels. The more often price responds to channel lines in the
t-grid, the more significant the t-grid becomes. This responding
to the channel lines develops and supports the notion of
the market being high (and attracting sellers) or low (and First, the bullish closing price reversal day respected the
attracting buyers). When a channel breaks then the channel resistance of the channel line. The next bar responded with a
lines reverse their roles—resistances become support and drop, but the narrowing and small range warned that the drop
supports become resistance. was of a minor corrective nature. The following three bars
provide the break of the channel line and its retest, suggesting
Progressing the t-grid that the psychology of the market was turning bullish. The
As a trend unfolds it will move through a series of channels. subsequent bar with an expanding range confirmed the market
In Figure 2 the Australian Dollar has broken a channel was at the lower end of the range and that a rise to the higher
resistance, suitably accompanied by a wider daily range, and end of the range likely which occurred in the next two bars.
sets up the opportunity to test the next channel line. On this Generally, once a channel resistance has been tested we would
occasion the rally stopped in the middle of the new channel expect a drop back to channel support. The last bar of Figure 4
and the last bar made a bearish closing price reversal. That bar is we see price action respects the resistance, but provides no
suggests downside action for the next bar. At this point the clear indication of direction. It will be the next bar that would
t-grid has not been confirmed, so we await the outcome of the confirm the direction.
drop back to the lower channel. Figure 3 shows the next couple Figure 5 shows the outcome:
of bars falling back into the lower channel. The following price bar pushed through the channel
The concept of resistance becoming support did not hold in resistance and closed at the high and above the previous bar’s
the strict meaning of this idea, and price returned into the lower high, providing a strong signal that the attempt to react to
channel. However, and of note, the downward momentum was channel resistance had failed and that once again the market is
not sustained as with the last bar prices moved up, producing a in a ‘low’ position. The next objective is the done by doubling of
bullish closing price reversal, a signal for higher levels to come. the channel; and as seen, the next solid line in Figure 5, which
The channel lines should be treated as a guide with the final was achieved several months later. In this case, the market
confirmation coming from price action. We now need prices did pause at the halfway point in the channel—the level of the
to follow-through on the upside and move back into the higher third channel in the sequence. Note how a closing price reversal
channel. signalled further downside. A contracting range provided some

PAGE 42 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

warning that the downside may not continue, but it was the Figure 6: Failure to resume the rally.
following month with its bullish expanding range that signalled
prices were ready to resume their ascent.

Figure 5: Another channel resistance breaks.

Trading the t-grid

By identifying key support and resistance levels within a
trend, the t-grid defines the market in temporary high or low
states that provide the opportunity for initiating trades and
closing them out. But to take action when confirmations are
provided by the market means that not all channels will be
The next channel resistance did not contain the rally and tradeable. Particularly in the early stages of the trend when
price continued to the next minor channel resistance. On then the range is narrow, the signals generated to buy and
this occasion there was no weakening of the uptrend as price sell do not offer a sufficient range to return a profit. As the
approached the channel resistance. There was no clear bar trend unfolds and the market trades over more than one base
signal for the top. Such action is a subtle sign that the market channel widths (as seen in the previous figures) will there be
remains in a strong underlying upward trend. an opportunity to profit from following the structure of the
The drop off the high did respect the support of the channel t-grid? To answer this, an easy measure to make is the ratio of
line, albeit for just the one bar, and at the time would have the channel width to the average daily trading range: the higher
generated expectations of the rally resuming. That expectation the ratio the greater the possible opportunity for a profitable
was squashed on the next bar as prices quickly gave way to a return. Where the ratio is small then the opportunity for profit
drop below the supporting channel line. The decline extended may be too small to justify the risk on the trade.
only to the next minor channel line, which coincides with the One way around the problem is to redraw the t-grid on a
level of support where the rally paused several months earlier. smaller time frame and follow signals generated there against
The rest of Figure 5 shows price contained within the the main channel lines to initiate and exit trades.
immediate minor channel just below the major channel As the trend evolves the channels become wider so in time
resistance. With the last bar the t-grid is neutral on trend the problem of the ratio diminishes.
direction. We can have a bias for the upside as price rose easily
through the main channel resistance on its way to its recent The Time Dimension
high, and since then has been responding to the upward sloping Just as key price levels are determined, key time points can
t-grid. However, the inability to hold above the main channel also be determined. Figure 7 shows the “vibrations” for the
resistance highlights current weakness in the trend. The t-grid Australian Dollar:
identifies the minor channel resistance that contained the
recent high as the level to break to clear the way for further Figure 7: time vibrations of the trend.
gains. A drop through the nearby minor channel support—that
contained the recent decline and marked a pause to the earlier
rally—will project prices back to the last major channel line
around 0.6300.
The t-grid is not a forecasting tool, although it has
forecasting implications; and merely highlights the key levels in
the market.
Figure 6 shows the next several months action:
Although the Australian Dollar did rise above the main
channel line price respected the minor channel line containing
the recent top. A bearish closing price reversal at the high, a
failed attempt to rally again and then a down bar, signals that
the attempt to take the market into a higher range is failing. The unit of vibration is the time between the two lows that
A drop back to the channel line that contained the last low may set up the t-grid. This unit can be projected forward and key
now be likely. price actions should be expected at these time points. Generally
the market will maintain a consistent theme over the period

IFTA.ORG PAGE 43
 IFTA JOURNAL 2012 EDITION

of the vibration—trending up, trending down or ranging. At another vibration cycle of building a base, price rallied again
times a combination will be seen such as trending for most of and quickly fell in with the existing t-grid structure. The steep
the vibration period but making a pullback into the end of the decline does not appear to have altered the t-grid. As a general
period. You should allow a “window” for the turn, generally one rule, only when the base trendline of the t-grid is broken can we
or two bars before and after the identified or projected date. accept that the t-grid is then invalid and expired.
Time is secondary to price and helps in understanding the
ebb and flow of the evolving trend. The time vibration is the Figure 9: Rally from the origin to the present.
additional information need to understand the position of the
market. It is particularly useful when price does not provide
decisive signals.
In Figure 7 we can see that the last high occurred within the
window for the end of the vibration period. Nevertheless with
the next bar, on the date for the vibration rollover, the drop was
arrested. The firm drop on the last bar confirms that a period
high is in place. Reviewing the last 2 vibration periods we can
see that price rallied during the first and built the top in the
second. Now with a high confirmed the expected outcome is a
decline in the next vibration cycle, to break the nearby channel
line and eventually test the minor channel line that contained
and supported the last sell-off. Figure 8 shows the outcome.
Later in 2010 prices returned to the channel line marking
the containment of the 2008 rally. After a vibration period
Figure 8: Outcome from the 2004 high
of testing this channel line prices moved above it in April
2011. This creates the opportunity of possible progression
to the next major channel line now near 1.5000. However,
and as we see, the move through the channel line has not yet
been confirmed; prices remain vulnerable to a breakdown
of support and a decline. While prices hold above the 1.0380
channel support, a rise through 1.1000 would confirm a new
rally. The Australian Dollar will then be ‘low’ again. Only a drop
below the 1.0300 channel line tell us the Australian Dollar is
vulnerable to a new decline.

Application
The t-grid divides the broader trend into smaller well-defined
The Australian Dollar took 2½ vibration periods to test the ranges. Such an approach will not suit all types of traders. Long-
minor channel line and along the way price responded around term traders who prefer a “buy and hold” approach may find
the immediate channel line. This action is a warning that price that the t-grid detracts from following the underlying trend.
outcomes are likely to vary from the expectations implied by However, the t-grid will still be useful as an analytical tool for
the t-grid. The action off the low eventually moved back above setting stops, adding to positions and re-entering trades.
the major channel line. Note how the market topped at the end The t-grid will best appeal to swing traders looking for
of the last vibration period, attempted a reversal that did not opportunities to trade defined ranges, buying once channel
break the immediate channel support. With the last bar now supports are confirmed and taking profit on tests of channel
closing at the high of the bar, the Australian Dollar appears resistances, and vice-versa.
ready to rally again. Traders can take advantage of redrawing the t-grid on
smaller time frames and take action closer to the moment
Back to the present channel lines are confirmed.
Let’s return to the Australian Dollar and see the result to the
present time (May 2011): Summary
We can see that throughout 2007 and into 2008 the The t-grid offers a way to get a handle on whether or not
Australian Dollar progressed into the next broader channel, the market is relatively high or low by breaking up the major
reacting around the minor channel lines and peaking with a trend into smaller ranges. It is not a full definition of the
bearish closing price reversal just before the next major channel trend and the system can expand indefinitely. As the trend
resistance. This time the market fell aggressively. The t-grid progresses volatility increases and the channel is continually
itself cannot anticipate such a reaction; it can only highlight doubled to account for that increased volatility. By waiting
potential levels of support and resistance and identify the for price action to confirm the validity of the channel lines the
rhythm of the move. Price fell through all but two channel lines, trader is presented with effective and well-defined trading
but the time of the decline spanned one vibration cycle. After opportunities.

PAGE 44 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

Is Average True Range a Superior Volatility Measure?

by Glenn Marci, CFTe, MFTA

Abstract To answer the second question, both concepts of volatility

This paper considers applicability of different volatility are scrutinized on a theoretical level before they are applied to
measures such as Average True Range and Standard Deviation. real markets. Differences of the measures are presented as well
During the recent financial crisis volatility has been difficult to as statistical analysis of real world data to make the measures
capture for a lot of technical indicators, so the task is to review comparable. For each of the measures a band system will be
the results of the two measures during the crisis. tested as well as a stop-and-reverse (SAR) system. Results will
During the course of the paper the question will be addressed be presented for different markets and market phases and the
whether or not there is a kind of “good” and “bad volatility”. Good change of volatility will be analyzed. In particular the impact
volatility would be characterized as volatility generated during of the financial crisis on the markets and the impact on trading
a trend phase and it should confirm a trend. Bad volatility would models will be evaluated.
therefore be rapid up and down swings in price movements,
where the market does not trend in a specific direction. 2. Theoretical Background
Therefore, a special emphasize will be put on crisis times as
during the past two years when prices in markets moved rapidly a) What is volatility?
and irrationally with intensive changes of trends. As a simple Volatility is a measure of a market’s variability. It could also
technical measurement tool, Bollinger Bands will be applied to be called a measure of a market’s activity. 3 There are very many
prices as opposed to a Keltner band system. In addition, a stop- ways to measure how dispersed prices on a market are, and
and-reverse system is employed for each of the measures. each focuses on a specific aspect of a movement or price range.
Volatility always depends on the time horizon on which it is
1. How Does Volatility Change During calibrated, but while some measures change dramatically with
Times of Market Crisis? time, others will only show moderate changes. Some measures
include all data points within the look back period; some only
Markets move in trends. This is one of the most fundamental use the extreme values during the time. However, all volatility
concepts in technical analysis and the concept is required for measures try to summarize the movement of a market or
acceptance to apply technical analysis. 1 A current market security over a specified time in one figure.
trend is more likely, over time, to continue than to end. As a Discernment has already been made between different
consequence, the question arises how far a trend will continue kinds of volatility. Is there good or bad volatility? Most
and when it is going to end. In order to capture these trends and technical indicators need a certain degree of price movement
underlying price movements there needs to be measures which to reach a meaningful level for which the indicator has a force
describe the magnitude and speed of a movement. Generally of expression. And most indicators try to give an indication
this is done via a concept that is described with the general term of the likelihood of prices in a specified direction. Either they
volatility. But what measures exist and what are the differences are confirming the direction of prices or trends as with using
among them? moving averages 4 or they are indicating that a reversal of the
Benoit Mandelbrot pointed out in his book, “The (mis) current price action is likely. An example of the latter type of
Behavior of Markets” that the magnitude of price moves often indicators would be the class of oscillators, which try to find
shows dependence to past moves, but that the direction of areas of overbought and oversold prices. 5
moves is not always clear. 2 He speaks of dependence without Here the focus is on volatility, generated during a trend
correlation as a strong fall in prices is more likely to be followed phase. It thus fits into the first category of indicators to
by yet another strong move, but the direction of the subsequent be confirming price action. As a consequence, volatility is
move may be uncorrelated to the past. Is there a possibility for a “good” if it rises due to price moves in one direction or due to
volatility measure to capture this kind of relationship? an acceleration of an ongoing price move, which is another
The first part of this study will answer the first question description of a trend. “Bad” volatility is then characterized
and review the basics of measuring price moves. To do this, a as volatility of prices which are fluctuating without trend
theoretical fundament is laid of the different possibilities of direction. This definition is close to the one given by Thomas
measuring volatility. Several volatility measures will be briefly Stridsman in his book “Trading Systems That Work”. 6 He called
touched, but the focus will be put on the concepts of standard volatility which increases with the direction of a trend “good”
deviation (SD) and average true range (ATR), together with their as it works in favour of positions from trend following systems
applications in technical indicators called Bollinger Bands and and everything that works against the positions from such a
Keltner Bands. system “bad”.

IFTA.ORG PAGE 45
 IFTA JOURNAL 2012 EDITION

Measures of volatility range from maximum-minimum with μ as average of the data during the past N periods.
comparisons, dispersion around a trend line, historical volatility John Bollinger uses SD in order to construct a measure of
or implied volatility as derived by options pricing to measure relative highness and relative lowness of prices by adding and
expected volatility in the future. Perry J. Kaufman suggests subtracting a multiple of SD from or to a moving average. 10
four measures of volatility which show the expansion and These so called Bollinger Bands are a well-known tool for
contraction of volatility over time, which would be a desirable examining price and trend developments. Together with other
feature in the analysis. 7 Besides price change over a defined indicators they provide a powerful tool for assessing if a market
time, maximum price fluctuation and the sum of absolute trend is more likely to continue than to reverse. Furthermore,
changes, Kaufman favoured the ATR as it is also suitable to SD can also be used to construct trailing stop systems, which
give an impression of future volatility. Thus, this measure has will adapt to the volatility of a market and give trading
already proven to have advantageous characteristics, so it will strategies enough room for drawdowns.
be the first candidate of volatility to be reviewed.
When John Bollinger was searching for a volatility measure c) Average True Range and its applications
to construct trading bands he favoured seven different volatility The concept of True Range and its application as a technical
measures and finally decided to use Standard Deviation (SD).8 indicator have been published by J. Welles Wilder jr. in his
In his view the special quality of this measure of volatility book “New Concepts in Technical Trading Strategies”. 11 It can
was to magnify the deviations from the mean of price. Thus be derived from the idea that a measure of volatility during a
SD will rise strongly when prices are far from the mean of a period could be defined as the highest value during the period
specified look back period. This is another way of saying that minus the lowest value. This trading range neglects the fact that
prices are trending in one direction. As it is our goal to detect so prices tend to jump from time to time so that the lowest value
called “good volatility” or trend volatility, SD is appealing as an of a period is still above the previous day’s market action. As a
adequate measure. result the idea of True Range (TR) arose, which is defined as the
greatest of the following:
b) Standard Deviation and its applications
SD is probably the most widespread measure for analysing 1. High–Low
how dispersion data. 9 There are many applications in 2. High–Previous Close
mathematics and statistics so it is very well explored and can be 3. Low–Previous Close
used to compare any kind of data. Together with assumptions
about the general distribution of data it helps to determine the To form an indicator of volatility TR is averaged over a
characteristics of a price distribution. It is defined as: specified period. Wilder recommended a period of 14 days and
defined a volatility measure called volatility index as:
N
2
∑(Pr ice i − m)
SD = i =1

€ƒSimple trend phase Source: Tradesignal, own data.

Figure 1:

PAGE 46 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

Figure 2: Trend reversal Source: Tradesignal, own data.

Figure 3: Trend reversal with gaps Source: Tradesignal, own data.

Figure 4: Rising price range Source: Tradesignal, own data.

IFTA.ORG PAGE 47
 IFTA JOURNAL 2012 EDITION

A simple average of the first 14 periods is used to calculate as defined by Welles Wilder with an exponential moving
the initial Volatility Index. The index is then calculated with an average. In this theoretical analysis, ATR with simple moving
exponential moving average of the previous true ranges. This average is added as it is mathematically and graphically more
volatility index has since been better known under the name of easy to see the effects.
Average True Range even though there can be slight differences The first set of price movements shows the change from a
in the calculation. Wilder used an exponential weighting while sideward phase to a trending phase. All candles are identical, so
most technical analysis packages now use a simple arithmetic the only change is in the direction of the movement. A change
average for the calculation of ATR. ATR and Volatility Index will in the direction of a price movement cannot be discovered by
be used synonymously in this paper. ATR, but has a severe effect on SD which rises until the whole
Wilder used ATR in a variety of ways. One of the applications indicator period comprises the uptrend. All indicators have a
was for a stop-and-reverse system, which deducted a multiple time horizon of 20 days so the effect on SD vanishes after 20
of ATR from the most favourable close during a trade he called days and SD stays on its higher level. From the chart it can be
“Significant Close.” This system was just named Volatility learned that the end of a sideward phase and the start of a new
System as it was the most basic application of the Volatility trend is accompanied by rising volatility. This means that e.g.
Index. Another very popular application of ATR is the Directional Bollinger Bands are expected to widen in such a setting. John
Movement Index (DMI), which measures the directionality of Bollinger notes that a sharp expansion of band width (meaning
a market by relating the biggest part of today’s trading range rising SD) from very low levels often marks “the beginning of a
which is outside of yesterday’s trading range against the ATR. sustainable trend”. 13
The calculations for this indicator will not be presented here, In the second setting a trend reversal is depicted with
but one interesting application of the system should at least be constant candles. ATR will not change in this environment. In a
noted. Wilders used an index called ADX (Average Directional first reaction SD will fall as a consequence of the trend change
Movement Index) to assess markets and their trend strength. as the extremes during the period under review move closer
The idea of trading ranges can be found from the very early together towards the average, which limits the square roots of
days of technical analysis. One of the popular systems which the deviations of these observations from their mean. Only after
take advantage of this concept is the Keltner Bands or Keltner half of the observation period (in our case 10 observations) has
Channels. They were introduced by Chester W. Keltner in 1960 in passed, volatility will rise again and finally reach its original
his book “How To Make Money In Commodities”. 12 He developed level in the end (after 20 periods).
trading bands which were based on a moving average with a Bollinger Bands would thus contract in reaction to a change
surrounding range constructed from what he called “typical of the prevalent trend, before widening again as the new trend
price.” This typical price was calculated as the average of the gains in range. In the real world such a setting is probably not
high, low and close price of the period. As in most band systems often as key reversals are often coincided by higher trading
the volatility measure was scaled by a factor (usually two) and ranges or small real bodies of candles, so ATR would be expected
added or subtracted from the mean. During the 1980s Linda B. to show higher readings. 14
Raschke modified the Keltner Bands and used ATR instead of the If a trend reverses and price gaps appear, it is often
typical price average and used an exponential moving average considered a critical phase for a price move. 15 Even though
instead of the simple moving average. Sometimes these band the candles in the chart above are unchanged, they are moved
systems are called Keltner Channels, sometimes Keltner ATR apart to show gaps, which result in a higher ATR. From the first
Bands. In later chapters an ATR system will be used, which is a day of the new trend, ATR rises and marks the change in the
combination of both developments. price structure. The first day of the new trend distinguishes
A simple moving average is used in order to receive comparable with a close benchmarked against the high of the uptrend. This
results, as from the Bollinger Bands and ATR—as defined above, is the reason why ATR rises up to the 21st day, before settling
this will be used as a volatility measure to construct the bands. on the 22nd day as it can be seen in the ATR calculated with the
During the course of the paper the constructed ATR band system simple moving average.
and Keltner Channels should be regarded as synonymous, The SD, however, once again leads the change in price
despite the minor differences in calculation. behavior as it adapts to the new environment. More generally
In the following section the behavior of the introduced it can be said as a rule of thumb, that SD will fall in the event of
measures of volatility, SD and ATR, are observed and compared. a trend reversal until the latest close in the opposite direction
Real world data is often too messy, so the start-off point is a of the previous trend exceeds all other closes during the look
simple data series that can be easily understood. References to back period.
applications of SD and ATR will be added to help readers recognize There is one last building block left to observe. Until now
specific patterns, which are visible in application of Bollinger we have not yet changed the candles themselves. At first the
Bands in the case of SD or the Keltner Channels for ATR. downtrend is weakening, so SD and ATR are both declining.
SD is showing much more of a decline as it reverts back to zero
d) Theoretical analysis from a very high level. If intraday variability picks up, then
The charts in this section are divided in four boxes with ATR shows this change at once, even if markets close just at the
the most upper one presenting p rice development, the second previous period’s close. This is an important insight as we can
showing the SD indicator, the third showing the ATR with a conclude that there is a fairly good chance that a breakout from
simple moving average and the lower box with the ATR indicator a trading range could first be visible in a rising ATR before it

PAGE 48 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

starts to show up in SD. Table 1: TR and SD as % of market value Source: Bloomberg, own calculations.
Thus a breakout should
first be visible in widening S&P DAX Bund Future Oil
Keltner Bands before Average 1.60% 2.12% 0.54% 2.63%
Max 6.93% 6.27% 1.02% 10.71%
Bollinger Bands will ATR (20)
Min 0.64% 0.78% 0.34% 1.28%
follow.
Max / Min 10.78 8.08 3.01 8.34
Another conclusion Average 1.88% 2.52% 0.60% 4.22%
that can already be drawn Max 12.72% 14.21% 1.74% 20.69%
is that SD will generally SD (20)
Min 0.39% 0.59% 0.16% 1.29%
be more volatile than ATR. Max / Min 32.85 23.96 11.03 16.00
While ATR is bound to
the trading ranges of the
market, SD depends on the closing prices and the distance from Model (1)
each other to rise. If ATR and SD are scaled to show the same One possibility is to apply ATR and SD in a band system as
mean then SD should show the greater variation for almost any it became popular with SD through John Bollinger’s Bollinger
market one could imagine. Bands or with ATR through Linda Bradford’s modification of
the Keltner Channels. For the Bollinger Bands the “most elegant
e) Real world data direct application” would be a volatility-breakout system. 16
This section gives a very brief impression of the differences It has to be noticed that this kind of system normally needs
of the two selected measures for the markets under review in prerequisites to give a trading signal. However, if one would
this study. The analysed markets, which will also be used for the apply such strict conditions, it would be very difficult to amass
evaluation of models based on SD and ATR, are: enough statistical data to gather a valid analysis, and trend
phases would be excluded.
ƒƒ Global Equities: As a consequence the trading bands systems will be used as
»» S&P 500 as a global they are, despite the expectable poor performance. Every move
benchmark index outside of one of the trading bands is followed until prices reach
»» DAX as a German index the opposite trading band. In this analysis it is not the goal to
ƒƒ Commodities: Oil prices measured by Brent oil (spot in order construct a system which yields a high profit or an optimized
to avoid rollover gaps) trading system, but to compare the two volatility measures in the
ƒƒ Fixed Income Markets: 10y Bund-Future (back adjusted) most pure way possible. The multiplication factor of SD will be kept
at 2.0 as suggested for the Bollinger Band system, while the factor
On the following pages (figures 5–8) a chart for each market for ATR will be subject to a scaling factor as explained below.
is shown with the market index as a black line, the SD as a blue
and ATR as an orange line. A red line separates the period before Model (2)
the financial market crisis from the crisis itself. The division A second approach will be used to validate the suitability
between these two periods is set on January 1st 2008. of ATR and SD as volatility measures. A system which is
Table 1 shows again that ATR is generally lower than SD on permanently in the market would be a good starting point as
average. Maximum and minimum values consistently vary more it gathers enough statistical data material to be of significant
for SD than for ATR. Both equity charts are highly correlated meaning. Once again Wilders offers a good starting point
and show similar percentage values for SD and ATR in table one. with his volatility system, which is a stop-and-reverse-
The oil market in figure 8 shows a very different picture: From system (SAR system) based on ATR. 17 It is a breakout system
a historic perspective SD and ATR both were relatively low, which does not rely on moving averages, but on distances
but exploded together with the oil price spike which began in from special points of the price history. From the “extreme
the middle of 2007. This is also reflected in the high maximum favorable close price reached while in a trade” a multiple of
percentage values of ATR and SD. All markets exhibit much ATR is added or deducted to construct a stop level for the
stronger values of ATR or SD during the crisis times and most trade. The prevailing trend is followed until the stop is hit and
record levels reached were a multiple of the values seen before. an opposite position is entered automatically. As this model
The Bund future chart is special as high fluctuations in SD and was designed for ATR the multiplication factor of ATR will
ATR are much more common during all time periods. Maximum be kept at the original default of 3 and SD will be adapted if
and minimum values are also rather close to the average. needed. The calculation period is set to 14 as in the original
ATR calculation.
3. Testing Setup
Volatility will rise during times of crisis as markets Adjustments
experience rapid change. Thus this study will have a focus As this study is on the different reactions to price and
on intense market moves as they occurred during the phase trend changes, it is necessary to make figures comparable.
between 2008 and 2009, when the financial markets crisis A reasonable way is to scale ATR and SD on the same level
affected almost every market. so that averages will be the same. Thus the pure effect of a

IFTA.ORG PAGE 49
 IFTA JOURNAL 2012 EDITION

Figure 5: S&P Index Source: Bloomberg data, own calculations.

140 1800

1600
120

1400

100
1200

80
1000
ATR / SD

S&P
800
60

600
40

400

20
200

0 0
01/00 01/01 01/02 01/03 01/04 01/05 01/06 01/07 01/08 01/09
SD 20 ATR 20 S&P

Figure 6: DAX Index Source: Bloomberg data, own calculations.

600 9000

8000

500
7000

400 6000

5000
ATR / SD

DAX
300
4000

200 3000

2000
100
1000

0 0
01/00 01/01 01/02 01/03 01/04 01/05 01/06 01/07 01/08 01/09

SD 20 ATR 20 DAX

different measurement method will not be tainted by different the best profit factors. Yet, it is intriguing that this is the only
levels of the volatility indicators. Details of this analysis can be market where both models performed relatively worse during
found in the appendix. the time of the crisis. An explanation for this can be deducted
from the findings of section 2e, where the Bund future showed
4. Findings the smallest variation of the measures over time and the
As expected, the following table (table 2) illustrates that the most consistent volatility pattern. Figure 8 shows that SD is
trading bands are very slow systems, which do not generate many fluctuating strongly around ATR with greater variation than in
trades. On average there were not more than 3 to 5 trades per the other pictures. This seems to be a favorable pattern for the
year. Bollinger Band system.
It is interesting to note that both models scored best for With the exception of the Bund market and the Bollinger
the Bund future market with the most profitable trades and system for the DAX, both models were better during the crisis.

PAGE 50 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

Figure 7: Oil Source: Bloomberg data, own calculations.

16 160

14 140

12 120

10 100
ATR / SD

Oil
8 80

6 60

4 40

2 20

0 0
01/00 01/01 01/02 01/03 01/04 01/05 01/06 01/07 01/08 01/09

SD 20 ATR 20 Oil

Figure 8: Bund Future Source: Bloomberg data, own calculations.

2,5 130

120
2

110

1,5
100

Bund Future
ATR / SD

90
1

80

0,5
70

0 60
01/00 01/01 01/02 01/03 01/04 01/05 01/06 01/07 01/08 01/09
SD 20 ATR 20 Bund Future

Volatility rose during the crisis and trends were very long in every market for the Keltner system. It is interesting that the
enduring. With regard to the profit factors from the two models, maximum profit of the ATR system occurred right in the times
no definite conclusion can be drawn from the data as to which of crisis for all markets. Bollinger Bands mostly yielded their
model performs best. Nevertheless there are some consistent biggest profit in 2008 to 2009, but not consistently, and the
differences between the models which are quite striking. The difference against the previous maximum profit is not as large as
ATR band system seems to be the more robust approach as the for the ATR system. Another fact of interest is that the Bollinger
average profit even rose during the crisis time for every single Band system shows more trades than the Keltner model. This
market. In most cases average profit even skyrocketed during applies to all time intervals under review. It appears that the
the time between 2008 and 2009 as, for example, in the S&P or Bollinger System is more sensitive towards the market resulting
the oil market. in trades being out more often than the Keltner system.
Maximum profits for all periods are also consistently higher Differences between the two models become clear if an

IFTA.ORG PAGE 51
 IFTA JOURNAL 2012 EDITION

Table 2: Trading Results of Band Systems Source: Tradesignal, own calculations.

Keltner Band System Bollinger Band System

Profit Factor

Profit Factor
% Profitable

% Profitable
# of Trades

# of Trades
Max Profit

Max Profit
Average

Average
Profit

Profit
2000-2009
all data 39 26% -0.22% 0.93 42% 48 29% 0.13% 1.06 32%
2000-2007
before crisis 37 24% -1.19% 0.63 27% 42 29% -0.16% 0.92 22%
S&P 2008-2009
during crisis 2 50% 17.60% 6.20 42% 6 33% 2.17% 1.60 32%
% Change
2008-2009 to
2000-2007 26% 18.79% 5.57 15% 5% 2.33% 0.68 10%
2000-2009 39 46% 2.04% 1.64 43% 48 48% 1.00% 1.34 37%
2000-2007 34 44% 1.02% 1.32 33% 39 49% 1.01% 1.36 37%
DAX
2008-2009 5 60% 9.00% 4.16 43% 9 44% 1.00% 1.26 26%
% Change 16% 7.98% 2.84 10% -4% -0.01% -0.10 -11%
2000-2009 30 60% 0.68% 1.85 10% 46 54% 0.58% 1.92 9%
2000-2007 26 65% 0.64% 1.94 6% 36 58% 0.66% 2.35 7%
Bund
2008-2009 4 25% 0.93% 1.61 10% 10 40% 0.27% 1.24 9%
% Change -40% 0.29% -0.32 4% -18% -0.39% -1.11 2%
2000-2009 35 34% 1.06% 1.17 94% 57 32% -1.77% 0.74 67%
2000-2007 28 29% -2.23% 0.66 55% 48 29% -3.40% 0.51 50%
Oil
2008-2009 7 57% 14.20% 3.38 94% 9 44% 6.93% 2.05 67%
% Change 29% 16.43% 2.73 38% 15% 10.33% 1.54 16%

Figure 9: Best trade of the Keltner bands on the DAX Source: Tradesignal, own program code.

example of a trade is reviewed. Figure 9 shows the most the market in November 2007. Considering the time to mid of
successful trade of the Keltner system on the DAX market. April the system was stopped out four times. This alone would
Starting on January 16th 2008, the model stayed short until not be too negative as stops also limit the risk of a trade. If
March 27th of the following year. From the chart it can be seen all trades during this period are summed up then they show a
that the ATR system is not overly volatile. Bands do fluctuate performance of 1462 points—less than half of the performance
in width and adapt to the markets, but have a very smooth of the Keltner system.
appearance. The short position yielded a gain of 3271 points. In figure 11 both band systems are plotted in the same chart.
In relation to the maximum price move, the system was able to The blue area shows the Bollinger Bands and the golden area the
recover 72% of the price range, an excellent result. Keltner Bands. Bollinger Bands show the greater variability, but
Interestingly the Bollinger Band system called to short this also means that they tend to narrow strongly if a market

PAGE 52 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

Figure 10: Bollinger system is whipsawed on the DAX Source: Tradesignal, own program code.

Figure 11: Keltner and Bollinger Bands Source: Tradesignal, own program code.

reverses quickly. This can be understood by the theoretical average winning profit. The logic behind this idea is that a stop-
analysis done in part two, which showed the behaviour of SD in and-reverse system is generally not efficient in detecting trends
case of a trend reversal. It is advantageous if a market reverses and often has drawdowns. It can be compared to the Parabolic
as the system is stopped out early on, but it seems that more SAR, also created by Welles Wilder. Research from Robert Colby,
often we should expect simply technical corrections from a though, suggests that the parabolic stop-and-reverse strategy
major trend which do not end in a new trend. in its original form is ineffective. 18 How to spot a new trend
Following are the results and review of the stop-and-reverse. exceeds the focus of this study. Ideas for this purpose can be
The focus now shifts from profit factors, which are important derived from many textbooks. 19 Focusing on winning trades,
for trading systems, to trades which prove to be favourable. the answer is given to the question: How effective is a trailing
In table 3 average profits are displayed, but our concern is the stop system based on either ATR or SD with a good entry point?

IFTA.ORG PAGE 53
 IFTA JOURNAL 2012 EDITION

Table 3: Trading Results of Stop-and-Reverse Systems Source: Tradesignal, own calculations.

ATR 14 SAR System SD14 SAR System

Average Profit

Average Profit
% Profitable

% Profitable
# of Trades

# of Trades
Max Profit

Max Profit
Winning

Winning
Average

Average
Profit

Profit
2000-2009
all data 92 40% 0.28% 5.14% 26% 134 34% -0.64% 3.67% 15%
2000-2007
before crisis 79 39% -0.08% 3.98% 17% 111 33% -0.57% 3.14% 15%
S&P 2008-2009
during crisis 13 46% 2.46% 11.12% 26% 23 39% -0.98% 5.85% 15%
% Change
2008-2009 to
2000-2007 7% 2.54% 7.14% 10% 6% -0.41% 2.71% -1%
2000-2009 88 41% 0.87% 7.61% 34% 134 39% 0.11% 5.64% 23%
2000-2007 72 39% 0.71% 7.60% 34% 110 38% -0.27% 5.03% 23%
DAX
2008-2009 16 50% 1.56% 7.66% 16% 24 42% 1.88% 8.23% 21%
% Change 11% 0.85% 0.07% -19% 3% 2.15% 3.20% -2%
2000-2009 76 43% 0.27% 2.00% 9% 128 46% 0.12% 1.40% 5%
2000-2007 62 45% 0.30% 1.81% 6% 98 50% 0.25% 1.34% 4%
Bund
2008-2009 14 36% 0.10% 3.04% 9% 30 33% -0.30% 1.67% 5%
% Change -9% -0.20% 1.23% 3% -17% -0.55% 0.33% 1%
2000-2009 137 40% -0.37% 8.58% 56% 213 35% -0.13% 8.25% 39%
2000-2007 115 37% -0.91% 7.57% 27% 169 34% -0.46% 6.89% 18%
Oil
2008-2009 22 55% 2.42% 12.24% 56% 44 39% 1.12% 12.79% 39%
% Change 17% 3.33% 4.67% 30% 5% 1.58% 5.90% 21%

Table 3 provides the results for the stop-and-reverse systems profitable trades with higher average winning profits.
build on ATR or SD. All models showed higher winning profits ATR seems to be a better and more stable measure for a
during the times of the financial crisis, but this does not imply market’s activity. Especially during the times of high volatility
that average profits generally rose during the crisis. As stop ATR proved to be a slightly more robust measure than SD.
levels are closer to the price action than in the band systems, Coming back to the question of good or bad volatility, ATR
trades are more frequent. Maximum profits occurred mostly appears to best adapt under conditions of extended trends.
during the crisis as for the band systems. This is another expression for the so called good volatility. In
For most of the markets the ATR system shows more sum the question, which is the title of this thesis, can generally
profitable trades and, generally, a higher average profit over the be answered with yes. ATR seems to outperform SD in many
total observation period. As already stated these outcomes are circumstances.
of minor interest, but interestingly the picture for the winning A general result of the study is that both volatility measures,
trades is clear: For all markets and all time periods except regardless of the model used, worked well during the financial
for one (oil during the crisis) the ATR system provided higher crisis. Volatility was at high levels and trends endured, but also
average winning profits than the SD model. with many whipsaws. Interestingly, the relationship between ATR
and SD remained stable during all time periods and in all markets.
5. Conclusion Further research might focus on the combination of these
In the first part of this thesis the differences between ATR two volatility measures to detect and exploit the advantages of
and SD were explained and the mechanics of the volatility each. Additionally, such an analysis would be most interesting to
measures were analysed. SD can be slow in the case of trend determine significant market turning points or breakouts from
reversals and often fails to detect new trends early on. ATR is sideward ranges.
much faster in the measurement of new trends, at least when
accompanied by higher trading ranges, as is typical for trend 6. Appendix
reversals as seen in figure 3. From this conclusion it is clear In tables 4–6, ATR and SD are compared during pre-
that ATR is best suited to situations of dependence of volatility crisis times (2000–2007) and within the major crisis time
without correlation, as described by Benoit Mandelbrot. ATR (2008–2009). There will be a comparison of the two measures
would take full effect of high activity markets, regardless for a 20‑day horizon (a usual time horizon for SD) and a 14-day
of their direction. These situations most often constitute a horizon (which is suggested by Welles Wilder for ATR).
precursor of a new trend. SD is lagging and could be described as On the 20 day horizon ATR is generally smaller than SD and it
a confirmation of a new trend. is interesting that this relationship is relatively constant during
The second part of this study supports the theoretical different points in time. With the exception of the oil market,
analysis. Maximum profits are consistently higher for ATR in the relationship appears much the same for different markets.
whatever testing model, for most markets using ATR show more If pre-crisis levels are compared with crisis levels, it can

PAGE 54 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

Table 4: ATR and SD characteristics for 20 day periods Source: Bloomberg, own calculations.

ATR (20 ) SD (20) SD / ATR

S&P 16.20 19.23 1.19
DAX 97.18 115.66 1.19
2000-2007
Bund-Future 0.51 0.57 1.13
Oil 0.98 1.59 1.62
S&P 25.58 28.94 1.13
DAX 144.82 170.17 1.18
2008-2009
Bund-Future 0.83 0.87 1.06
Oil 2.53 4.12 1.63
S&P 57.9% 50.5% Average ATR Change
% Change 2008-2009 to DAX 49.0% 47.1% 82.2%
2000-2007 Bund-Future 63.7% 52.7% Average SD Change
Oil 158.2% 160.1% 77.6%

Table 5: ATR and SD characteristics for 14 day periods Source: Bloomberg, own calculations.

ATR (14 ) SD (14) SD / ATR

S&P 16.21 16.48 1.02
DAX 97.18 97.95 1.01
2000-2007
Bund-Future 0.51 0.48 0.94
Oil 0.99 1.36 1.38
S&P 25.47 24.79 0.97
DAX 144.85 144.00 0.99
2008-2009
Bund-Future 0.83 0.75 0.91
Oil 2.53 3.49 1.38
S&P 57.1% 50.4% Average ATR Change
% Change 2008-2009 to DAX 49.1% 47.0% 81.6%
2000-2007 Bund-Future 63.8% 57.5% Average SD Change
Oil 156.5% 157.1% 78.0%

Table 6: ATR and SD scaling factors for the models Source: Bloomberg, own calculations.

SAR System Band System

ATR (14 ) SD (14) ATR (20 ) SD (20)
S&P 3 3 2.30 2
DAX 3 3 2.30 2
Bund-Future 3 3 2.30 2
Oil 3 2.14 3.20 2

be seen that both measures of volatility react in almost the limited than deviations from the mean of a longer period. For
same manner. Both rise about 50% if compared to pre-crisis the examined period of 14-days, ATR and SD are almost on
levels. The oil market is forming a specialty as it showed rapid the same level. This is one of the first interesting results as
price changes during the crisis. In consequence, volatility as it suggests that the 14 day period is a good basis for relative
measured by both measures rose about 160%. comparisons of ATR and SD.
ATR remains almost unchanged when the look back period is During the further course of this paper the standard period
shortened. This is not overly surprising as every True Range is for the two volatility indicators will be set at 14 days for the
included in the calculation and averaged afterwards. Whether Volatility SAR system and at 20 days for the band systems.
the 14-day average of a measure or the 20-day average of a While the 14 day measure do not need adjustment, except for oil,
measure is taken, it influences only the speed of the indicator the ATR of the band systems will be adjusted by a factor of 2.3
reactions. SD on the other hand is much lower if the period is instead of 2.0 as used for SD. Measures of ATR will be scaled by
shortened. Also this result is not very surprising as deviations 1.38 for the oil market, so the average deviation will be on the
from the mean during a shorter period of time will be more same level as SD.

IFTA.ORG PAGE 55
 IFTA JOURNAL 2012 EDITION

7. References
1. Murphy, John J., Technical Analysis of Financial Markets, New York
Institute of Finance, New York, 1999, p. 3.
2. Mandelbrot, Benoit B. and Richard L. Hudson, The (Mis)Behavior Of
Markets, Basic Books, New York, 2004, pp. 247-248.
3. Wilder, J. Welles Jr., New Concepts In Technical Trading Systems,
Hunter Publishing Company, Winston-Salem, 1978, p. 21.
4. Murphy, John J., Technical Analysis of Financial Markets, New York
Institute of Finance, New York, 1999, p. 195.
5. Murphy, John J., Technical Analysis of Financial Markets, New York
Institute of Finance, New York, 1999, p. 227.
6. Stridsman, Thomas, Trading Systems That Work, McGraw-Hill, New
York, 2001, p. 99.
7. Kaufman, Perry J., New Trading Systems and Methods, John Wiley &
Sons, Inc., Hoboken, 2005, p.796.
8. Bollinger, John, Bollinger on Bollinger Bands, McGraw-Hill, New York,
2002, pp. 51-52.
9. Kaufman, Perry J., New Trading Systems and Methods, John Wiley &
Sons, Inc., Hoboken, 2005, p.28.
10. Bollinger, John, Bollinger on Bollinger Bands, McGraw-Hill, New York,
2002, p.52.
11. Wilder, J. Welles Jr., New Concepts In Technical Trading Systems,
Hunter Publishing Company, Winston-Salem, 1978, p. 21-26.
12. Keltner, Chester W., How To Make Money In Commodities, The Keltner
Statistical Service, Kansas City, 1960.
13. Bollinger, John, Bollinger on Bollinger Bands, McGraw-Hill, New York,
2002, p. 67.
14. Edwards, Robert D. and John Magee W.H.C. Bassetti, Technical
Analysis Of Stock Trends, 9th ed., Amacom, Boca Raton, 2007, p.181.
15. Pring, Martin J., Technical Analysis Explained, 4th ed., McGraw-Hill,
New York, 2002, p.106.
16. Bollinger, John, Bollinger on Bollinger Bands, McGraw-Hill, New York,
2002, p. 127.
17. Wilder, J. Welles Jr., New Concepts In Technical Trading Systems,
Hunter Publishing Company, Winston-Salem, 1978, p. 23-26.
18. Colby, Robert W., The Encyclopaedia of Technical Market Indicators,
McGraw-Hill, New York, 2003, p. 496.
19. Murphy, John J., Technical Analysis of Financial Markets, New York
Institute of Finance, New York, 1999, p. 385.

Bibliography
Bollinger, John, Bollinger on Bollinger Bands, McGraw-Hill, New York,
2002
Colby, Robert W., The Encyclopaedia of Technical Market Indicators,
McGraw-Hill, New York, 2003
Edwards, Robert D. and John Magee W.H.C. Bassetti, Technical Analysis
Of Stock Trends, 9th ed., Amacom, Boca Raton, 2007
Kaufman, Perry J., New Trading Systems and Methods, John Wiley &
Sons, Inc., Hoboken, 2005
Keltner, Chester W., How To Make Money In Commodities, The Keltner
Statistical Service, Kansas City, 1960
Mandelbrot, Benoit B. and Richard L. Hudson, The (mis)Behaviour of
Markets, Basic Books, New York, 2004
Murphy, John J., Technical Analysis of Financial Markets, New York
Institute of Finance, New York, 1999
Pring, Martin J., Technical Analysis Explained, 4th ed., McGraw-Hill, New
York, 2002
Stridsman, Thomas, Trading Systems That Work, McGraw-Hill, New
York, 2001
Wilder, J. Welles Jr., New Concepts In Technical Trading Systems, Hunter
Publishing Company, Winston-Salem, 1978

Software and data

Model building is done with Tradesignal Enterprise Edition, Tradesignal
GmbH, Bremen, Germany
Tables and some charts are created with MS Office XP Standard Edition
from Microsoft Corporation, Redmond WA, USA
Data for this study is provided by Bloomberg

PAGE 56 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

The Volatility-Based Envelopes (VBE): a Dynamic

Adaptation to Fixed Width Moving Average Envelopes
by Mohamed Elsaiid, MFTA

Abstract Adopting the statistical concept of standard deviation to

This paper discusses the limitations of fixed-width envelopes the field of technical analysis, Mr. Bollinger introduced the
and introduces a new method that addresses these limitations. Bollinger Bands (B-Bands). The B-Bands are two bands set at
The new method utilizes the concepts of standard deviation two standard deviations above and below a SMA calculated
and correlation to produce a dynamic adaptation to the fixed- off the price action. In principle, the B-Bands aim at utilizing
width envelopes. The paper also offers an example of a useful the standard deviation concept in order to identify rare and
technique and some guidelines for applying the new method on unsustainable price excursions and coin them as overbought
price charts. The method will be referred to henceforth as the (OB) and oversold (OS) conditions.4
volatility-based envelopes (VBE). The B-Bands manage to contain more price action within
its boundaries, especially during trendless phases in price
Introduction action where identifying OB and OS conditions using the
B-Bands become quite valuable. However, there are certain
Fixed-Width Envelopes price conditions on the near to short term horizon as explained
Fixed-width envelopes (FWE) are two boundaries. Each is by Bollinger, in which prices tend to breakout and remain
placed at a fixed percentage above and below a simple moving outside either one of the 2-standard deviation bands for
average (SMA) of the exact same duration. The primary aim of some considerable time. At other times, even if the price
using the FWE is to contain the price action fluctuations and, excursion was relatively brief, the price gain (or loss) would
hence, imply when prices have become over-extended in either be considerable. These conditions will generally occur during
direction. FWE are characterized by the same effects of lag and trending phases and following periods of low volatility in price
smoothness associated with their corresponding SMA.1 action and are dubbed by John Bollinger as volatility breakouts.
Unfortunately, due to the inherent lag effect caused by The technique proposed by Bollinger relies on these volatility
the FWE, prices would quite often move and remain outside breakouts in order to initiate a position in the direction (favor)
the envelopes’ boundaries for a notable period of time. of the price breakout.5 Though very successful when properly
Despite some featured techniques adapted for the FWE that identified in the price action, these volatility breakouts seem
would accommodate and sometimes even depend on these to argue against the general notion that price excursions
occurrences2, the initial purpose to contain the price action is occurring beyond two standard deviations are deemed rare
not satisfied. and unsustainable.
Bryan J. Millard suggested that in order to represent the Sustainable price breakouts from the B-Bands are
trend properly and highlight active and dominant cycles using primarily attributed to the difference in tendency and
SMAs and FWE respectively, the statistically-correct plot behavior of price action during trending vs. non-trending
would be to shift it back from the most recent data point by phases. During non-trending phases, price action visually
half the span of the SMA duration. This technique is referred exhibits a characteristic of oscillatory/mean reversion
to as centering the moving average. The rationale behind this motion. During these events, price excursions are rare and
is that since the FWE are properly plotted (centered), the price unsustainable. While during trending phases, this feature
action fluctuations will be contained within the envelope becomes less dominant and further diminishes on the near
boundaries.3 to short term horizon as the trending phase grows stronger.
To its credit, the centered FWE manages to contain a larger This is attributed to the lagging effect of the SMA which
amount of price action. However, it still produces challenges. visually appears clearer during trending phases. As you
As a result of the centering procedure, the envelopes’ values would recall from the B-bands calculation, the 2-standard
will terminate n-days prior to the most recent closing price, deviations calculated are added to and subtracted from
where n = (the SMA span – 1)/2. Moreover, the centered FWE that lagging SMA to construct the upper and lower bands
are non-adaptive to the continuous volatility changes of respectively. Hence, the B-Bands do not fully resolve the lag
the price action. Depending on the price volatility, this will effect of the SMA.
frequently cause the price fluctuations to move and remain out Although the B-Bands succeed in achieving adaptability,
of the envelope boundaries (during high volatility phases), or the upper and lower bands do not inherit the smoothness of
even not react with the envelopes at all (during low volatility their corresponding SMA as they are relatively more erratic
phases). in motion than the latter. This does not allow them to be as
Another attempt to address the FWE’s lack of adaptability suitable as centered FWE when attempting to highlight active
to price volatility was made in the 1980s by John Bollinger. and dominant cycles in the price action.

IFTA.ORG PAGE 57
 IFTA JOURNAL 2012 EDITION

Introducing the adaptive Example:

Volatility‑Based Envelopes (VBE) Assuming the following data:
The last given price (S) of the NASDAQ Index is: 2,190.
Using volatility to achieve adaptability
To address the drawback associated with the lack of The simple average (μ) of the percent change is: 0.07%.
adaptability of the FWE, we use the measure of standard
The (σ) of the daily percent change is: 1.00%.
deviation. Unlike B-Bands’ calculation, we use the historical
percent changes of price returns of a security instead of the Therefore, we can expect that approximately 95.4% of the
historical price action of that security. daily percent change movements to be maintained within the
Practitioners in the field of statistics and financial percentage range of:
engineering have hypothesized over the past decades that the
percent changes in a stock price (or security) are normally 0.07% – (1.00% * 2) = – 1.93% (at 2 standard deviation).
distributed on the short term.6 Hence, we use this hypothesis
0.07% + (1.00% * 2) = + 2.07% (at 2 standard deviation).
as the basis for the VBE calculation methodology; once the
standard deviation calculations were complete, the outcome
To translate those values into a price range for most recent
was added to and subtracted from a SMA of the percent
closing value of the index, or in other words, the raw VBE, then:
changes of price returns. Then, the outcome was added to/and
subtracted from today’s (the most recent) closing value on a 2,190 * (1 – 1.93%) = 2,147.7 (lower raw envelope at
percentage basis and not over a lagged SMA of the price action. 2 standard deviation).
As a result, a dynamic adaptation to the envelopes’ boundaries
can be achieved, while avoiding the inherent lag effect of the MA 2,190 * (1 + 2.07%) = 2,235.3 (upper raw envelope at
of prices. The following steps will explain the VBE’s calculation 2 standard deviation).
methodology:
Plotting the raw VBE over the price chart
Step 1: Calculate the standard deviation (σ) of the percent Using the same calculation method presented above, we can
changes (or logarithm) of the daily historical price returns (σ). regress and calculate a daily range for all previous historical
closing values of the S&P 500 Index and then plot the outcome as
The standard deviation is calculated over duration of 21 daily shown in Figure 1.
percent change values. Figure 1 depicts the S&P 500 Index line chart with daily
closing values and the raw upper and lower boundaries of the
Step 2: Calculate the values of the raw Volatility-Based calculated raw VBE. As observed, there exists a strong (almost
Envelopes (raw VBE). identical) similarity between the closing values (line chart) of
the S&P500 Index and both the upper (red) and lower (blue)

Figure 1: S&P 500 Index – Line chart – Daily closing values – Normal scale

PAGE 58 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

boundaries of the calculated raw VBE. Having said that, both volatility of the S&P 500 Index movements. Meanwhile, unlike
boundaries are choppy (raw), just as the index movement. Thus, the B-Bands, the VBE maintains its boundary smoothness,
a need to smooth out these boundaries is required. relative to the corresponding moving average of the price action.
And finally, the VBE managed to contain more price action than
Step 3: Smooth the raw VBE using weighted moving averages the B-Bands. The VBE is constructed with the primary advantage
To smooth out the raw VBE, we will use two centered of its ability to identify overbought (OB) and oversold (OS)
weighted moving averages (CWMA) for both envelopes of the conditions in the price chart regardless of the trend status.
raw VBE. Using CWMAs instead of CSMAs mathematically
results in a reduction of lag-time by approximately 40%. Step 4: Forecast the VBE’s missing data points using
This means that instead of lagging the most recent price by correlation
(span  -  1)/2 as with the case of the SMA, the lag is reduced to be To forecast the missing data points of the VBE, we use both
approximately equivalent (span - 1)/3.34. 7 the CWMA feature previously presented in Table 1, as well as
In real life observations, and mainly due to the non-linear the statistical concept of correlation (ρ). The aim is to use the
nature of price action, the lag tends to be reduced down to equate correlation between the values of other CWMAs of lesser span
(span - 1)/4 instead of (span - 1)/3.34. This means that— in real life (independent variables) with the 21-period CWMA or smoothed
price action—the lag of the 21-period WMA tends to approximate VBE (dependant variable) to forecast the missing data points of
to 5-periods (and in some cases, 4-periods), but not 6-periods. that smoothed VBE. It’s worth mentioning that all CWMAs of
lesser span are selected with reference to the amount of their
The smoothed Volatility-Based missing data points.
Table 1 features
CWMAs of different Envelopes (VBE)
spans and the amount of Now let us make a visual Example:
lag attained by each. comparison between the smoothed Using the daily values of the S&P 500 Index, we calculate (ρ)
VBE vs. both the centered FWE and the matrix of the daily percentage change of a 21-day CWMA vs. the
CWMA lag
B-Bands. This comparison is shown in daily percentage change of a 17-day, 13-day, 9-day, 5-day and
span periodic Figure 2. a 2-day CWMA over the most recent 63-actual data points as
21.00 5.00 Figure 2 depicts the advantages of shown in table 2 (below).
4.00 the VBE over the centered FWE and
17.00 Table 2: S&P 500 Index – correlation coefficients (ρ) of 21, 17, 13, 9, 5
the B-Bands. The centered FWE failed
13.00 3.00 and 2-day % change of CWMAs
to mechanically adapt to volatility
9.00 2.00 changes during the movements of the
21-CWMA 21-CWMA 21-CWMA 21-CWMA 21-CWMA 21-CWMA
1.00 S&P 500 Index, while the VBE was able
5.00 21-CWMA 1 0.88 0.76 0.67 0.56 0.31
to contract and expand in accordance
2.00 0.25 to the decrease and increase in

Figure 2: S&P 500 Index – Candlestick chart – Daily closing values – Normal scale

IFTA.ORG PAGE 59
 IFTA JOURNAL 2012 EDITION

Using (ρ) to forecast the missing data points of the . Using that same concept, we can now forecast the five missing
21-day CWMA data points of the smoothed VBE.
As previously explained, the 21-CWMA has 5 missing data Figure 3 depicts the smoothed VBE (at 2-standard deviation)
points, while the 17-CWMA has only 4. This means that we can with a forecast of its missing five data points using the
use the last given value of the 17-CWMA and the (ρ) value of correlation methodology previously presented.
both variables from table 2 to forecast the 1st missing value of
the 21-CWMA as follows: Using the VBE to identify over-extended
price action on the price charts
Example:
Referring to the data used in calculation, the last calculated Now that the VBE has been constructed, we will demonstrate
percent change of the 17-CWMA was 0.80%. The last calculated a useful trading technique when applying it to price charts.
value of the 21-CWMA was 1,122.30. The calculated (ρ) value was Below are some essential guidelines to be followed when using
0.88 or 88% (from Table 2). the VBE.

Then, the forecast of the 1st missing value of the 21-CWMA ƒƒ Spot the most recent turning phase of the VBE (crest or
would be: trough) while it is occurring. The turning phase must be
associated with a price excursion. The VBE will guarantee to
1,122.40 * [1 + (0.80% * 0.88%)] = 1,130.38. a high degree that any price excursions are unsustainable
regardless of the trend.
This value is placed shifted back from the most recent closing ƒƒ If a price excursion occurred at a low, wait for the price to
value of the index by 4-days (since the 17-CWMA has only 4 return back inside the VBE range, and then initiate a long
missing data points). position (or buy-back an old short position) until the next VBE
turn (in the opposite direction) takes place.
Moving onwards, the following table (Table 3) shows the last ƒƒ If a price excursion occurred at a high, wait for the price
calculated percent changes of the 13, 9, 5 and 2 CWMAs as well to return back inside the VBE range, and then short, sell or
as the forecast of the 2nd, 3rd, 4th and 5th (i.e. last) missing values reduce your position until the next VBE turn (in the opposite
of the 21-CWMA: direction) takes place.
Table 3 Needless to say, the appropriate trading strategy applied will
13-CWMA 9-CWMA 5-CWMA 2-CWMA depend on the direction of the overriding trend direction.
0.86% 1.07% 0.93% 0.80% The following example (Figures 6 and 7) illustrates how to
1,137.80 1,145.99 1,151.90 1,154.77 initiate buy and sell trades using the VBE.

Figure 3: S&P 500 Index – Candlestick chart – Daily closing values – Normal scale

PAGE 60 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

Figure 4: EGX 30 Index – Candlestick chart – Daily closing values – Semi-log. scale

Figure 5: EGX 30 Index – Candlestick chart – Daily closing values – Semi-log. scale

Figure 6: NASDAQ Index – Candlestick chart – Daily closing values – Semi-log. scale

IFTA.ORG PAGE 61
 IFTA JOURNAL 2012 EDITION

Conclusion References
The VBE introduced in this paper dynamically adapts to the 1 Murphy, John J., Technical Analysis of the Financial Markets, New York
institute of Finance, 1999.
volatility changes of the price action and thus, successfully 2 Pring, Martin J., Technical Analysis Explained: The Successful
contains the price action within a predefined standard deviation Investor’s Guide to Spotting Investment Trends and Turning Points,
range. Accordingly, the VBE is consistently able to identify McGraw-Hill, 2002.
[3, 7] Millard, Brian J., Channels and Cycles: A Tribute to J. M. Hurst,
over-extended price action regardless of the trend status. Traders Press, 1999.
This is achieved without compromising the smoothness of its [4, 5] Bollinger, John A., Bollinger on Bollinger Bands, McGraw-Hill, 2001.
boundaries. 6 Hull, John C., Options, Futures, and Other Derivatives, Prentice Hall,
2000.
Nevertheless, the VBE is still left with a few challenges. Most
importantly, is the fact that the most recent data points on the
smoothed VBE are missing and required a forecast. In this paper, Bibliography
we used the concept of correlation and applied it to moving Hull, John C., Options, Futures, and Other Derivatives, Prentice Hall,
2000.
averages of different durations in order to achieve a reliable Mason, Robert D., Marchal, William G., Lind, Douglas A., Statistical
forecast for the missing data points. Still, the correlation figures Techniques in Business & Economics, McGraw-Hill/Irwin, 2002.
tend to lose their significance as they approach zero, since a Millard, Brian J., Channels and Cycles: A Tribute to J. M. Hurst, Traders
Press, 1999.
value of zero implies no correlation between the variables. Thus, Bollinger, John A., Bollinger on Bollinger Bands, McGraw-Hill, 2001.
the significance of the VBE estimated values will vary depending Murphy, John J., Technical Analysis of the Financial Markets, New York
on the significance of the correlation figures, which tend to institute of Finance, 1999.
Pring, Martin J., Technical Analysis Explained: The Successful Investor’s
change more often than not. Thus, one should always check the Guide to Spotting Investment Trends and Turning Points, McGraw-Hill,
(ρ) matrix values for statistical significance (i.e. at least above 2002.
0.5 and/or below -0.5). VIX White Paper, Chicago Board Options Exchange (CBOE), 2009.
Data courtesy of Bloomberg and Reuters.
Charting software courtesy of Equis International MetaStock v.9.1.

Figure 7: NASDAQ Index – Candlestick chart – Daily closing values – Semi-log. scale

PAGE 62 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

Alpha Trading: Profitable Strategies That Remove

Directional Risk by Perry Kaufman
Reviewed by Larry Lovrencic

The world has experienced many financial crises, from the South Sea bubble in the
early 18th century, to the Great Depression in early 20th century, to the Dot Com Crash
in the early 21st century, to the housing and credit crisis in more recent years. It was
the latter two crises which prompted author Perry Kaufman to consider alternative
financial markets trading strategies that require more complex positions other than
just long or short, but which he suggests rewards the trader with safety during a
market collapse. The result of Kaufman’s deliberations is the book Alpha Trading:
Profitable Strategies That Remove Directional Risk.
According to Kaufman the focus of this book is on the active trade, and the trading
strategies employed are called Statistical Arbitrage or StatArb, which is the trading
basis of many of the large hedge fund managers throughout the world. Statistical
arbitrage is a market neutral trading method where a trader takes advantage of the
expected mean reversion of the relationship between two co-integrated securities.
Effectively, by cancelling out Beta (the overall market’s contribution to a security’s
return) and trading Alpha (the active return of a security), a trader can feel confident
and expect positive returns during extremely volatile periods of market activity. In
this book, Kaufman takes the reader on a journey through formulae and examples to
illustrate effective methods for achieving trading returns with lower risk profiles.
The book begins by contemplating the importance of price noise in a market
with Kaufman examining the “drunken sailor walk” and the efficiency ratio, and
quickly moves to a substantial section on the process of trading pairs of securities—
initially equities and then futures. Examples are thorough and provide a good basis
to understanding the process. Sample spreadsheets are provided on a website link
referred to in the book.
In the following chapter Kaufman examines longer term pairs trading where price
noise is usually not dominant and where well defined trends tend to emerge. The
relationships between Dell and Hewlett-Packard and gold and platinum are used as
examples. Cross market trading is also examined with a focus on the use of the “stress
indicator”.
Alpha Trading is similar to Kaufman’s previous books, New Trading Systems and
Methods and A Short Course in Technical Trading, in its examination of the quantitative
aspects of trading, but it departs by not using Technical Analysis as the focus of
the strategies. It provides well explained, robust methods for traders considering
alternative quantitative trading strategies.

IFTA.ORG PAGE 63
 IFTA JOURNAL 2012 EDITION

The Evolution of Technical Analysis

by Andrew W. Lo and Jasmina Hasanhodzic
Reviewed by Regina Meani, CFTe

Having reviewed The Heretics of Finance for the 2010 issue of the IFTA Journal it
was with great anticipation that I began reading this work by the same authors. While
written from an entirely different slant, this new endeavour did not disappoint.
As the name suggests the authors investigate the path of technical analysis through
the epochs, but the book reveals far more than this. History lovers amongst us, they
will not be let down as Lo and Hasanhodzic succeed in tracing the origins of technical
analysis back to the ancient Babylonian times, leading most of us on a fascinating path
of new ideas and discovery.
When chronicling and signposting the stages in the developments for technical
analysis, parallels are drawn to historical breakthroughs in civilization. The legendary
and great contributors to the technical analysis are the markers along the way. This
lends the book a reference quality as Lo and Hasanhodzic quote authors of some of
these fabled works, providing the novice with an excellent starting overview, and
reference points for their technical studies.
It is beyond the purpose of this book to cover the techniques and methods of
technical analysis in depth. Nor do the authors claim to be technicians in their own
right. And as an aside, this may be why the authors refer to technical analysis as a
“craft” in this historical overview of Technical Analysis instead of an intellectual skilled
discipline excercised with crafts and tools of the discipline. Lo and Hasanhodzic do
present an objective view of the great debate on the validation of technical analysis.
They track the history of technical analysis, but go much further and investigate and
propose possible holes in the Efficient Market Hypothesis (EMH) and the Random Walk
Theory. “A growing number of finance academics are coming to recognise that efficient
markets are not an adequate model of reality. Thus, a crack in the door has been opened
for academic considerations of technical analysis”. 1
The authors go on to suggest that Lo’s “adaptive markets hypothesis offers an
internally consistent framework in which the EMH and behavourial biases can
coexist” 2 and that the implications of such are significant for technical analysis and its
credibility.
In essence, The Evolution of Technical Analysis passes one of the most important tests
as it motivates its readers to find out more on this fascinating subject.
The review copy was provided courtesy of John Wiley & Sons Australia and organised
through the Educated Investor Book shop, Melbourne, Australia (see advertisement on
page 40).

Endnotes
1. A W Lo and J hasanhodzic, The Evolution of Technical Analysis, John Wiley & Sons, New Jersey, 2010, p.153
2. Ibid, p.164

PAGE 64 IFTA.ORG
 IFTA JOURNAL 2012 EDITION

Author Profiles

Robin Boldt, MFTA, SAMT John Gajewski

Robin is a Sales Trader for Swiss Equities at UBS Investment John Gajewski has 25 years experience in technical analysis
Bank, Zurich. He began his career in the field of Technical and the foreign exchange, fixed interest and commodity
Analysis in 2002 at JRC Capital Management in Berlin as a markets. At the Commonwealth Bank of Australia he produced
Technical Analyst before he joined Goldman Sachs’ one delta the successful Chartpoints commentaries, providing analysis
trading team in Frankfurt until 2009. and recommendations on forex, bonds and commodities to
As a sales trader Robin successfully implements Technical the bank’s traders and sales staff, institutional, corporate
Analysis not only to improve his execution quality but also and business clients. He now produces The Chart Manager for
embeds different tools of the Technical Analysis within his Stafford Blue Pty Ltd.
equity recommendations. And despite the fact that trading John has also been an active participant in technical analysis
and execution quality has become more and more dependent education, working with the Australian Professional Technical
on electronic algorithmic trading capabilities, he is fully Analysts Association, the Financial and Securities Industry
convinced that Technical Analysis gives traders and investors a Association and Kaplan Professional in developing and running
meaningful advantage. technical analysis courses.
Breakouts are of particular interest for him. His work aims
not only at the profitability of price channel variations, but also Regina Meani, CFTe
reminds of the fact that other parameters surrounding the core Regina covered world markets, as technical analyst and
trading strategy influence the trading profitability significantly. Associate Director for Deutsche Bank, before freelancing. She
is an author and has presented internationally and locally and
Mikael Bondesson lectured for the Financial Services Institute of Australasia
Mikael Bondesson, born 1977, holds a Degree of Master (FINSIA), Sydney University and the Australian Stock Exchange.
(One Year) of Science in Business and Economics with a major She is President of the Australian Professional Technical
in Economics from the Lund University, Sweden. He currently Analysts (APTA) and immediate past Journal Director for IFTA.
works as a sales trader and advisor within Private Banking Regina carries the CFTe designation. She has regular columns
at Ålandsbanken Sverige AB, where he advises private clients in the financial press and appears in other media forums. Her
and smaller institutional investors. His work also includes freelance work includes market analysis, private tutoring,
producing technical analysis within the bank; holding seminars webinars and larger seminars, advising and training investors
and workshops; training private clients, institutional investors and traders in Market Psychology, CFD and share trading
and colleagues in technical analysis. Mikael is a member of the and technical analysis. Regina is also a past director of the
Scandinavien Technical Analysts Federation (STAF). Australian Technical Analysts Association (ATAA) and has
belonged to the Society of Technical Analysts, UK (STA) for over
Mohamed A. Elaasar, MFTA thirty years.
Mohamed Elaasar is the Senior Technical Analyst and Market
Strategist since 2003 at EFG Hermes Holding, the largest
investment bank in the Middle East, and the cofounder of its
Technical Analysis Department. He is currently responsible
for issuing multiple types of technical research and reports
covering Egypt and Gulf markets in the Middle East. He works
with HNW clients, and provides technical training courses for
the company’s staff. He was previously responsible for the US
market analysist.
In 2009, he was grandfathered into the Society of Technical
Analysts (STA) in UK.
He began his career in the field of Technical Analysis in 1995
at Pan Arab Investment in Cairo as a Technical Analyst till 2001.
The scope of his work was mainly on global financial markets,
including forex, commodities, etc.
Mohamed appears frequently in regional financial media
channels and publishes articles about markets outlook.

IFTA.ORG PAGE 65
 IFTA JOURNAL 2012 EDITION

Board of Directors
The International Federation of Technical Analysts, Inc.

President Directors at Large

Adam Sorab, CFTe, MSTA (STA) David Furcajg, CFTe, MFTA (AFATE)
Vice-President – the Americas Akira Homma, CFA, CIIA, CFTe, FRM (NTAA)
Timothy Bradley (TSAASF) Julius de Kempenaer (DCTA)
Peter Pontikis (STANZ)
Vice-President – Europe
Antonella Sabatini (SIAT & SAMT)
David Sneddon (STA)
Vice-President – Asia Staff
Taichi Otaki (NTAA)
Executive Director
Vice-President – Middle East, Africa
Beth W. Palys, FASAE, CAE
Mohamed Ashraf Mohfauz, CFTe, CETA (ESTA)
Vice President, Meetings
Treasurer
Grace L. Jan, CMP, CAE
Michael Steele (AAPTA)
Member Services Manager
Secretary
Saleh Nasser, CMT (ESTA) Linda Bernetich

Education Director (Academic & Syllabus) Senior Graphic Designer

Rolf Wetzer, Ph.D. (SAMT) Jon Benjamin

Accreditation Director Production Manager

Roberto Vargas, CFTe (STA) Penny Willocks
Exam Management Director Accounting
Gregor Bauer, Ph.D. (VTAD) Dawn Rosenfeld
Journal Director
Regina Meani, CTFe (STA, ATAA) The International Federation of Technical Analysts, Inc.
9707 Key West Avenue, Suite 100
Membership Director Rockville, MD 20850 USA
Dan Valcu, CFTe Email: [email protected]
Conference Director Telephone: +1-240-404-6508
Elaine Knuth (SAMT) (Immediate Past IFTA Chair) Fax: +1-301-990-9771

PAGE 66 IFTA.ORG
 IFTA Certified Financial
Technician (CFTe) Program
The IFTA Certificate (Certified Financial Technician)
consists of CFTe I and CFTe II, which together
constitute a complete professional program.
The two examinations culminate in the award of this
internationally recognised professional qualification in
Technical Analysis.

Examinations
The exams test not only technical skills, but also
international market knowledge.
CFTe I: This multiple-choice exam covers a wide range
of technical knowledge and understanding of the
principals of Technical Analysis, usually not involving
actual experience.
The CFTe I exam is offered in English, French, Italian,
German, Spanish, and Arabic, and is available, year-
round, at testing centers throughout the world, from
IFTA’s computer-based testing provider, Pearson VUE.
CFTe II: This exam incorporates a number of questions
requiring an essay based analysis and answers. For
this, the candidate should demonstrate a depth of
knowledge and experience in applying various methods
of technical analysis. The exam provides a number of
current charts covering one specific market (often an
equity), to be analysed, as though for a Fund Manager.
TThe CFTe is offered in English, French, Italian,
German, Spanish and Arabic bi-annually, typically in
April and October.

Curriculum
The program is designed for self-study. Local societies
may offer preparatory courses to assist potential
candidates. Syllabus and Study Guides are available on
the IFTA website at http://www.ifta.org/certifications/
registration/

To Register
Please visit our website at http://www.ifta.org/
certifications/registration/ for registration details.

Cost
IFTA Member Colleagues Non-Members
CFTe I $500 US CFTe I $700 US
CFTe II $800* US CFTe II $1,000* US

*Additional Fees (CFTe II only):

$250 US translation fee applies to non-English exams
$100 US applies for non-IFTA proctored exam locations
PICTURE
PERFECT
Perfect your trade strategies with charting
on Bloomberg. CHART <GO> is your
visual gateway to 20 million securities,
fundamentals, economic releases & more.
All this, integrated into an intuitive charting
platform with technical alerts, market moving
events, custom studies, backtesting and
impressive visualizations to boot.

Explore CHART <GO> on the BLOOMBERG For more information about Bloomberg,
contact a Bloomberg Sales representative
PROFESSIONAL® service. or visit www.bloomberg.com.

New York +1 212 318 2000

London +44 20 7330 7500
Tokyo +81 3 3201 8900
São Paulo +55 11 3048 4500
Dubai +971 4 373 9400

©2011 Bloomberg Finance L.P. All rights reserved. 44756261 0711