Selection of Best Players in ODI Cricket using

Ranking Based Indexing Method

MSc Research Project

MSc Data Analytics

Bhaskar Reddy Yenuga

Student ID: x21150567

School of Computing
National College of Ireland

Supervisor: Cristina Hava Muntean

 National College of Ireland

MSc Project Submission Sheet

School of Computing

Student …….…………Bhaskar Reddy Yenuga……………………………………………………………

Name:

Student ID: ………………………x21150567…………………………………………………………………….……

Programme: ……MSc Data Analytics…………………………… Year: ……2022-23….

Module: ……………MSc Research Project……………………………………………………….………

Supervisor: ……………Cristina Hava Muntean………………………………………………………….………

Submission
Due Date: …………21-12-2022…………………………………………………………………………….………

Project Title: ….Selection of Best Players in ODI Cricket using Ranking Based
Indexing Method

Word Count: …………21………………… Page Count……………7515……………………………….…….

I hereby certify that the information contained in this (my submission) is information
pertaining to research I conducted for this project. All information other than my own
contribution will be fully referenced and listed in the relevant bibliography section at the
rear of the project.
ALL internet material must be referenced in the bibliography section. Students are
required to use the Referencing Standard specified in the report template. To use other
author's written or electronic work is illegal (plagiarism) and may result in disciplinary
action.

Signature: ……………Bhaskar Reddy Yenuga………………………………………………

Date: ………………21-12-2022………………………………………………………………

PLEASE READ THE FOLLOWING INSTRUCTIONS AND CHECKLIST

Attach a completed copy of this sheet to each project (including multiple □

copies)
Attach a Moodle submission receipt of the online project □
submission, to each project (including multiple copies).
You must ensure that you retain a HARD COPY of the project, both □
for your own reference and in case a project is lost or mislaid. It is not
sufficient to keep a copy on computer.

Assignments that are submitted to the Programme Coordinator Office must be placed
into the assignment box located outside the office.

Office Use Only

Signature:
Date:
Penalty Applied (if applicable):
 Selection of Best Players in ODI Cricket using
Ranking Based Indexing Method
Bhaskar Reddy Yenuga
x21150567

Abstract
In an effort to gain priority, cricket players are rated to show their superiority over
rivals to acquire a position. Numerous scholars have proposed various statistical
techniques to evaluate teams and their players in studies on this topic. The ability to
compare a player's bowling and batting results have been established. In contrast to
previous studies, it is difficult to achieve the goal of player and team selection. In the
context of One-Day International (ODI) cricket, the research focuses on the prediction of
the best player's selection for team-making using a ranking-based indexing method based
on linear programming. Additionally, the dataset contains information from the
ESPNCRICINFO website by web scrapping about seven important features, namely
batting, bowling, fielding, partnership, all-rounder, batting from first innings, and
bowling from first innings records. All the characteristics and several attributes are well
analysed through the linear discriminate method to obtain the rank of each player and
lastly the final rank for the selection of the 11 best players from a team bench of 15.
Each rank of the cricketer's role for the selected players suggests which player plays
which team's role along with their final rank estimation. In comparison, it has been found
that 9 players from the predicted players from selection were selected for one ODI match
played against Australia. It is calculated that the accuracy rate of the findings is close to
82%.

1 Introduction
Without using the best performance indicators available, selecting a cricket squad cannot
be done reasonably. The conventional bowling average can be wildly inaccurate, particularly
when there are few runs scored but a large percentage of them are not outs. The current study
focuses on selecting a team after a limited number of games have been performed using the
most appropriate metrics. It is anticipated that game circumstances may affect the rate at
which batters score. To choose a squad from the athletes that competed in the ODI Series for
the International Cricket Council (ICC) Champions Tournament, these criteria are utilized as
an example. The use of a mixed integer linear technique in the selection procedure is
demonstrated (Shah 2017). The idea is that a cricket team should consist of a variety of
specialties, including wicketkeepers, all-rounders, bowlers, and batters. Ranking batters by
batting aptitude and bowlers by bowling capability may be possible for a selection panel, but
ranking all-rounder is more difficult. The ranking all-rounder is challenging because they
must be both a good batter and a decent bowler by necessity. In addition, how numerous of
every category of representative should be chosen? Integer performance tuning, a
scientifically valid approach, is utilized in this research to demonstrate how cricket squad

1
selection might benefit from its application. There are also recommendations provided for
choosing a squad if professional measured values must be used (Singla and Shukla, 2020).

International games, Limited - overs games, and ODI games are some of the forms used
in these sports. In addition to this, there are several major tournaments performed at the club
and county levels around the nation. An appropriate squad of performing 11 and 4 extra
individuals must be chosen to create a side for any such competitions, series, or World Cup.
One wicketkeeper with the ability to bat makes up a cricket team together with a group of
batters and bowlers. The team captain and selectors must choose the bowlers and batsmen for
the squad, along with a specialist batsman. The squad has a variety of spinners, including
fast, moderate fast, and spin spinners, and each hitter is trained to bat at a certain position
within the playing eleven. Based on the opposition team the team is competing against, team
members must be chosen from a group of players (Prakash, Patvardhan and Singh, 2016).
The chosen team may also be examined for its ability to outperform or set goals higher than
the team they are playing. Therefore, choosing such a squad is prejudiced and prone to error;
this choice is made based on the individual's expertise and analytical skills when analyzing
player key metrics and data.
Managers and captaincy may choose the most effective squad with the best
expectation of winning the match by automating the recruitment process by considering
quality metrics and predicting the match outcome. The absence of data for new players who
haven't participated in many international series might be another obstacle during choosing.
In this case, it is crucial to compare the new player's numbers from his club matches and a
few international matches to those in the player dataset already in place (Vetukuri, Sethi, and
Rajender, 2021). In this, a team selection technique and game consequence prediction tool
with player characterization capability to systematize this procedure and provide the new
player's set of metrics. Numerous individuals have researched squad selection, team
placement, and cricket game victory predictions. The 15-member cricket team selection
process is automated using optimization algorithms (11 players and 4 additional players: A
total of 15 players for a team bench). This method considers individual players as a bit more
information in a sting, representing the squad as a thread. These characteristics are later
turned into fitness by applying an optimal solution to assess them. To choose the Indian
World Cup team, this variation was created by cross-over, recombination, and substitution
processes (Perera 2021).
Utilizing a brief cricket tournament, a team is chosen using linear mathematical
programming. The measurement that is produced by contrasting team players with the team's
success in the series is suggested after a detailed discussion of the multiple measures utilized
in the research. The best batsmen order and match result prediction are carried out either
before or after the team is chosen to project the team's success (Jha et al. 2022). By
considering, the cricket and bowling traits of the chosen players, an ideal team composition is
generated using the stochastic optimization process from the vast combinatorial field. By
combining a supervised and unstructured technique, closest neighbor grouping and linear
regression are utilized to predict the result of the one-day worldwide game. One of the key
elements in forecasting the result of the match is the maximum number of runs that will be
scored in it, which may be anticipated using these approaches. To make predictions,
historical information gathered from prior matches is merged with current match data, such
as the number of catches and batting average. While the game is taking place, the logistic
regression algorithm is utilized to extract characteristics from the one-day cricketers,
drastically reducing the variables (Jha et al. 2022). The characteristics that must be utilized
for the models are chosen using the passing approach. It is possible to use multiple models to
forecast the performance of the team batted first along with the win margins in ODI since the
2
scores achieved by each side and the differential between the points of the two groups are
roughly distributed according to the normally distributed. Using the match situation at the
beginning of the session, the pre-match team quality, and other aspects of test cricket,
multivariate logistic modeling is used to estimate the result of the test series as a multinomial
answer like win prediction, loss prediction, and draw prediction (Jayanth et al. 2018).

Problem Statement - Who will be picked to represent each country's squad as a player?
Cricket is a game that needs a lot of human judgments; thus, it is tried to account for as many
aspects as it can be used to estimate the best player choice for a game. The main factors
affecting a player's performance are their dependability and batting/bowling percentage. To
predict players, the researcher used both their overall numbers and current form. The
approach has been improved by incorporating numerous different aspects that impact the
selection of the players in a team and the formation of a team for a specific match, even
though in earlier references metrics were used to forecast the winning potential after carefully
examining the results of the variables that influence the Player’s selection for a match.

Aims and Objective -

The research's clear objective is to choose the top 15 players for the squad that can prepare
for international competition. Finding out all important and irrelevant aspects that contribute
to the performance gaps between players is exceedingly difficult. However, it should be
overcome through clever reasoning and prediction models based on their success rates. So, to
accomplish this goal, one should think about the following goals, which are given below.
1. To gather and construct the database from the reputed website like as ESPN for the
Indian country.
2. To visualize the database as a feature importance
3. To show the best possible permutation and combinational approach by linear
programming
4. To measure the player's performance by similarity index and rank method.
5. To rank each player before the final selection of a player.
6. To estimate the final rank of the selected 15 players to pick up the best 11 players.

Research questionnaire-
• How to collect and gather the dataset construction from records from websites?
• To rank the players, player performance metrics are quantified using database
information. How to give more priority to the player's ranking?
• How to forecast the match before starting from the historical records of both teams'
players?
• How to select the final squad from the 15 players for the national team to play in
international games.

The document goes as follows, after introducing the research, it follows with some literature
review, methodology used to work on the motto of the project, implementation, evaluation of
the work, then some cases which show the working of the code, then followed by the results
and finally concluding the report with the conclusion and future scope of the project.

3
2 Related Work
Very few articles about players' prediction problems in the cricket game were found after
research. The effectiveness of cricket players has only been the subject of a very modest
number of studies. The effectiveness of Indian bowlers versus seven professional players,
which even the Indian cricket squad faces the most, was forecasted by the researcher. To
forecast how many scores a bowler is likely to give up and the number of wickets a bowler is
inclined to take in a specific ODI match, researchers used backpropagation networks and
nonlinear activation network functions (Vetukuri, Sethi, and Rajender, 2021). A multilayer
linear model was used in some papers to forecast the effectiveness of cricketers in a test
series. A standard for comparing and choosing batsmen in restricted over cricket was
established by researchers. They employed a two-dimensional visual illustration, with a
Scoring Rate on one dimension and chances of getting out on the other, to construct a new
metric. The selection process is then established based on the batsmen's striking rate and runs
scored. The researcher employed neural network models to forecast player performance,
categorizing bowlers and batsmen into three categories: performance, middling, and
unsuccessful. They suggest whether a player can be selected in the roster to participate in the
World Cup based on how frequently the player has gotten different ratings (Chakraborty,
Kumar, and Ramakrishnan, 2018).
The researcher analyzes the characteristics of the two sides to forecast the result of a
cricket match. They calculated this by evaluating the individual player abilities of each squad.
Researchers created techniques to simulate the abilities of bowlers and batters, and these
algorithms evaluate a player's ability by first analyzing his lifetime record and then the most
current achievements. The researcher developed a brand-new metric called the Cumulative
Bowling Average to evaluate bowlers' effectiveness. The bowling averages, scoring rate, and
efficiency are three conventional bowling metrics that make up the total bowling rate. This
combined striking rate was to evaluate the effectiveness of the batsmen in the cricket
tournament (Bhattacharjee and Saikia, 2016).

2.1 New measurement on Player's evaluation

The new measurement for bowlers also takes into consideration the caliber of each
batsman they are bowling too, as well as the caliber of each bowler they are facing. The
overall performance index of a batsman is the sum of the individual performances of the
batsman against each delivery. The overall quality index of a pitcher is the sum of that
bowler's performances against each opponent (Shah and Patel, 2018). For developing
algorithms to forecast the results of matches, the investigation established the terms "batting
index" and "bowling index" to rank individuals' performances. Using mathematics, the
proposed method was evaluated in the meaning of the ideal batted sequences for ODI
matches. How statistical simulation might be employed in predictive analysis for various
sports was discussed. The related paper reviewed the data mining techniques that had
previously been utilized to forecast sports outcomes and outlined the benefits and drawbacks
of each approach. The proper investigation was through the multiple records of the cricket
players who employed machine intelligence methods to forecast football game results. Neural
networks were employed to choose players (Premkumar, Chakrabarty, and Chowdhury
2020). The research was likely the first generalized method to forecast the number of runs

4
batsmen will score and the number of wickets a player will take on a specific match day. It is
impossible to generalize for all spinners worldwide because it was restricted to eight Indian
players and used neural networks to estimate how so many wickets a spinner will achieve. It
has developed a predictive prototype that may be used to forecast every player's success in a
particular match using machine intelligence algorithms.
With the emergence of online social networking sites and databases, academics'
attention has been drawn to the finding, ratings, and predictions of subject matter experts or
contents. The academic world has proposed several efforts to solve these challenges. Among
many other complex outcomes, locating Young Stars in scholarly networking, predicting
future reference counts, and finding historical experts are some of these proposals. But the
sports realm was not considered in these recommendations. Although several concerns with
sports prediction are examined while taking the sports domain into account, the researchers
do not propose a suitable prediction system. Additionally, none of the plans took cricket into
account. The observer determined the top bowlers and batters based on a cricket match.
Additional rankings of batsmen based on their productivity metrics and rankings of cricket
teams using h-index and Page Rank have been proposed. The following suggestions,
meanwhile, did not provide any Young Stars prediction framework for the sports industry
(Hussain et al. 2019).
In cricket, evaluating a player's performance is essential for creating a well-balanced
squad. Since playing circumstances vary from stadium to stadium, different tours necessitate
various player groupings. Thus, in addition to certain other factors like a player's experience,
performance in each circumstance, and various other characteristics, selectors must take into
account a player's numerous personal traits. One may get this data from a player's overall
career (Bhattacharjee and Saikia, 2016). To estimate the value of the qualities of the bowlers
and batsmen in the given game and aid in the selection of players for the specified tour, this
article discusses the ideas of various unrelated forest regressions. The ODI version of the
tournament will employ the concept. The dataset used to create the model is a player's prior
performance versus a certain opponent. The model considers the visiting team, the opponent,
and the match location. Using the input fields, a rank-wise table of all the batters and bowlers
is produced, which the assessors may use to choose the squad according to the preferred
mixture.
In the batting and bowling fields, measures are devised specifically to predict coming
stars. More specifically, three categories—Co-players, Team, and Opposing Teams—are
included, and 9 and 11 attributes, respectively, are established for the forecast of coming stars
in batting and bowling. Based on performance evolution measures and weighted averages,
two different types of datasets are produced. Utilizing both generating (BN and NB) and
discriminatory (SVM and CART) computational methods, the defined features are evaluated.
Co-batsmen dominate the other two categories in the batting area, while Team performs best
in predicting budding stars in the bowling field. Ultimately, it is shown that NB works better
than the other models. Finally, a rankings list of young talents for both domains is offered
based on the weighted average, performance progression, and rising star score (Prakash,
Patvardhan and Singh, 2016). These rankings are contrasted with the ICC ratings for 2013–
16, and it is discovered that the methods, it was provided are useful for predicting rising stars.
Therefore, in the ODI and T20 formats, similar traits may also be utilized to forecast rising
stars. To get even better outcomes, extra characteristics like opposing team variety, whether
home or away, 100s, the 50s for batsmen, and 4, and 5 wickets for bowlers may also be
included. Identifying RSs in cricket and other fields is very helpful since it allows the
authority like coaches, and managers, to focus their efforts on maximizing the knowledge of
such rising stars to achieve the best achievements in the future (Jha et al. 2022).

5
 The prediction of young stars may be implemented using a similar concept in a wide
range of sports organizations. The concept for win forecast and team model construction,
player assessment, player-specific information gathering, and player performance verification
are the four steps of the conceptual methodology for match win predictions, team assessment,
and player suggestion. The fragmented matching information is pre-processed and recorded
in the data storage during the first stage. The player productivity characterization model
utilizes the information of the players recorded in the system to measure and classify the
individuals. This information is provided at this stage. The following two steps make use of
these player metrics and player characterization information. The player measurement and
historical match win or loss data are utilized to develop the SVM for forecasting the victory
or defeat percentages in the win forecasting and team structure assessment phases. In the
concluding stage, the preferable role for a particular player is suggested using the grouping
and k-nearest neighbor approaches (Premkumar, Chakrabarty, and Chowdhury 2020).

2.2 Past related works

S.No. Reference Objectives and Outcomes Critical Analysis
1. Singla, S. and A prospective method of squad selection Integer Linear
Shukla, S.S., that suggests a Dream 11 Fantasy squad programming method
2020. for the forthcoming match utilizing real- and Markowitz
world information gathered from Player Optimal solutions are
achievements in the previous 10 matches. used. As a result of the
The Python Gurobi package was used to cricket team selection,
construct the Integer Linear programming there is no
method. The Markowitz Optimal solution, consideration for the
which is often used to choose stocks for rising players and new
financial investment, has also been applied players’ chances in the
to study the squad selection issue. playing 11 from the
team squad.
2. Jayanth, S.B., The player rating score is calculated by the Here predicting game
Anthony, A., investigator utilizing game and player score is specially
Abhilasha, G., information taken from a certain mentioned but there is
Shaik, N. and competition. The n-dimensional no mention of how a
Srinivasa, G., information taken into consideration for new player can show
2018. modelling cannot be sequentially the effect on the
separated, according to test findings. As a highest player’s
result, the asymmetric SVM with RBF performance.
kernel performed better than the linear and
multimodal kernels. The accuracy,
specificity, and recall rate of SVM with
RFB kernel are 75, 83.5, and 62.5
respectively. So, for predicting game
outcomes, it was advised to use SVM with
the Kernel function.
3. Hussain, A., The abilities and flaws of regional teams Here there is no
Qiang, Y., play a crucial role in choosing the teams. mention of failure rate
Bilal, A.Q.M., To rank cricket teams in the area, and combination of old
Ullah, U. and researchers in this work presented the and new players even
Ullah, N., Region-wise Team Rank (RWTR), an selection of the best
2019. adaptation of the PageRank method. It team. Page rank
6
 makes sense to award more credit to a algorithm is used here.
squad that defeats a tougher opponent. This algorithm is used
Compared to the standard rating, the for quick search
suggested ranking more accurately techniques.
captures the gameplay's flow.
4. Hussain, A., To rate cricket teams, they suggested a This paper is most
Yan, Q., Bilal, Region-wise Team Rank (RTR) and a suitable compared to
M.A.Q., Wu, Region-wise Weighted Team Rank other papers. But
K., Zhao, Z. (RWTR). The logic is that a squad that partnership records are
and Ahmed, defeats a better team should receive more not considered here. If
B., 2019. scores than a team that defeats a weaker considered, then the
team and vice versa. When determining a player’s ranking and
team's ranking score, the suggested team selection might
technique considers a team's area-specific be different.
strengths and weaknesses in addition to
the number of wins and losses in that
region. In closing, the teams' rankings are
compared to the actual ICC rankings.
5. Jha, A., Kar, In this essay, a two-step process for In the paper, the
A.K. and selecting teams and evaluating players in player’s ranking was
Gupta, A., fantasy cricket is proposed. Sequential not made by all the
2022. backward reduction in the random forest is necessary features. But
used to evaluate individuals while the sequential
considering situationally player statistics backward reduction in
and a revised evolutionary method for the random forest is
player choice. The outcomes demonstrate used here. It is very
that the suggested strategy outperforms the useful for each
conventional hit-and-miss approach used player’s comparison in
in fantasy athletics to choose teams. depth.
6. Premkumar, Here, new factors and improvements to Here factor analytical
P., current variables have been incorporated method is used to
Chakrabarty, based on specific productivity indicators, determine the factor
J.B. and which will determine a player's ranking in score but there are
Chowdhury, cricket. The former online leaderboards, missing steps for the
S., 2020. notably the most popular International player’s analysis in
Cricket Council scoring system, depth.
overlooked many of these factors. This
article evaluates batsmen and bowlers
who've already played One Day
International (ODI) cricket throughout the
calendar year 2015 using a dynamic
instead of a static technique of calculating
factor scores using the factor analytical
method, on a match-by-match premise.
7. Agrawal, P. The selection procedure for the Indian Integer optimization is
and Ganesh, cricket team's ODI roster is described in used but no proper
T., 2020, this essay. The examination of many attributes like player’s
April. criteria that might be used in the selection partnership and all-
process is provided. The rankings of the rounder fields are
reliable and unreliable players are missing for

7
 described in this examination. The justification.
participants will be picked in a range of
categories, including wicketkeeper,
spinner, and batsman. The player's
constancy was the major factor considered
in the rating. Coding for integer
optimization has been used for that
measurement.
8. Shah, P. and In this study, they used Principal A comparison between
Patel, M.N., Component Analysis to rank captains batsman and bowler
2018. according to several criteria. Additionally, was made but no other
the author has added a weighted average role was mentioned.
way to evaluate captains depending on the
team's rating scale (Z score), as well as the
captain's skill as a batsman and bowler.
9. Prakash, C.D., Throughout this study, a novel indicator Only DPI was based
Patvardhan, C. called the Deep Performance Index (DPI), on bowlers and
and Singh, S., which represents the effectiveness of the batsmen.
2016. batsmen and bowlers on a closer
examination of the needs of T20 cricket, is
created using a network having to learn
methodology. When developing the DPI,
significant characteristics are extracted
and their respective value is determined
using the Recursive Feature Elimination
technique, which is based on learning
algorithms. It is demonstrated that, when
contrasted to other well-known ranking
factors for T20 cricket, DPI is better
capable of collecting achievement
information for both bowlers and batsmen.
10. Chakraborty, Separate rankings are first created from Isolate ranking was
S., Kumar, V. the extensive dataset usually containing made but no actual
and the performances of various Test evaluation for ODI
Ramakrishnan, cricketers, comprised of controllable cricketers.
K., 2018. figures of applicant alternative solutions
while trying to impose some restrictions
on the least number of overs did play (for
batsmen), a limited number of tests did
play (for wicketkeepers and pacemen), as
well as the minimum figures of points
scored, and pitches chosen to take (for all-
rounders). Later, the TOPSIS approach is
used to rate those bowlers who had made
the summary and determine which players
will perform well in the projected World
XI Test squad.
Table 1: Related work

8
3 Research Methodology
The dataset was collected from the website ‘ESPN Cricinfo’, so it fully belongs to
secondary data analysis. Web scraping coding is one of the best choices for selecting the data
directly from the website. There is a reduced percentage chance of an error occurring while
fetching data from the live websites. In the project, analysis has been done for the data
collected by this method. A total of 7 datasets are collected from the ESPNCRICINFO
website. Some of them are in different lengths of dimension. The datasets are selected by
querying from the main website and the span of the dataset is up to the last 4 years i.e., Dec
2018 to Dec 2022. Most datasets are in 50 rows with different attributes. Each dataset is
described in the below section in table format. In the below section, they have 50 rows each
up to 4 sections, player-partnership with 446 rows, and 62 rows for the last two sections.

Table formats of dataset: Note that: (for setting weight)

M H VH L
Moderate High Very High Low

i. Batting Section
Player Span Mat Inn NO Runs HS Ave BF SR 100 50 0 4s 6s
H H H M VH M H VH H H

ii. Bowling Section

Player Span Mat Inn Overs Mdns Runs Wkts BBI Ave Econ SR 4 5
H H VH H H H VH VH

iii. All-rounder
Player Span Mat Runs HS Bat_Av 100 Wkts BBI Bowl_Av 5 Ct St Ave_Diff
H H VH VH M H L H H H

iv. Fielding
Player Span Mat Inns Dis Ct St Ct_WK Ct_Fi MD D/I
H VH VH VH H H H

v. Player-partnership
Player1 Player2 Wkt Runs Overs RR In Out Inns Opp Ground Start
Date
VH VH H H

vi. Batting in 1st innings

Team Score Overs RPO Inns Result Opposition Ground StartDate

vii. Bowling in 1st innings

Team Score Overs RPO Inns Result Opposition Ground StartDate

9
The methodology followed here to analyse and predict the squad for the country in ODI
is KDD (Knowledge Discovery Data). At the beginning of the data assessment, there are
many more steps required to apply in the data pre-processing steps.
Data Pre-processing Steps:
• Data quality assessment - In this step, a check has been done in the dataset if there
are any missing values from the dataset. Also, the analysis has checked the data types
whether the data is an integer, float, or any string category.
• Data Cleaning - In the next phase mainly data errors like non-responsive and
observational errors have been checked and repaired or transformed into the data for
analysis. Here empty values were changed into zero values.
• Data Transformation - In this phase new fields have been estimated by a bunch of
numerical techniques. Here feature importance, aggregation, and averaging methods
are adopted.
• Data reduction - In this phase, the dataset has been evaluated as in short dimension,
so it is called "Data reduction". Some attributes are selected to manipulate the player's
ranking.
• Important Feature Selection by weight-based aggregation – In every dataset of
cricket’s characteristics there are a lot of attributes for identifying the player’s role
and performance on the Cricket Ground. But say essentially not all attributes are
equally important to use for finding the player’s evaluation or selection purpose. And
even from all selective attributes it is not required to set equal priority for finding
ranking weightage. By based on setting asymmetric weight values to the selective
features player’s rank was made through this method.

Figure 1: Methodology applied on the data (KDD)

4 Design Specification
After collecting whole datasets to select the final 11 players, it is essential to estimate the
ranks for each player from all available records. How rank is estimated in linear
programming fashion and what is benefitted against the different classifier’s accuracy rate –
which is important to find out the novelty of the work with compared to past related works.
The advantage of this method over the use of different classifiers is maintaining the static
result and not looking for other variation results by the unequal classification rate. Hence, it is
a big advantage for the linear discriminate ranking method. In this section, the analysis has
been described briefly.

10
Proposed Diagram:

Figure 2:Proposed methodology for selection best players done by linear programming method.

5 Implementation
Selecting one cricket player from the large set of collections is not made easier because there
are a lot of attributes that talk about the cricketer’s habit on the ground. Which attribute is
importantly observed for the player’s selection – it is vital for giving the highest weightage
(to the attribute). Henceforth the team selectors never look at all attributes symmetrically.
Hence, weight multiplication for finding the total weight is required. The following linear
equations are essential to find out the rank of each player.

represents the comparative weights’ calculation based on selected featured

importance. represents the respective weights of the selected attribute ‘i’.
represents the value of represents the similarity index
(estimated by mean calculation of attribute i).

11
 represents the rank of the player. This rank estimation is the subset of final
ranking set.

represents the total weights of selecting final rank.

represents the final rank of selected players.

The following steps are described for the rank analysis as well as team selection.
• Step 1: Internet surfing to find the data about the Indian cricketers.
• Step 2: Data query from the website for finding the records of batting, bowling,
partnership, all-rounder, and fielding, etc.
• Step 3: Data collection about the 7 type datasets.
• Step 4: from every characteristic find the similarity index by calculating the average
value of the whole index and subtracting between each player's record and average
value. It provides positive as well as a negative values. the positive value means an
upwards difference from the average and the negative value means vice versa. After
this finding, each player's rank is estimated.
• Step 5: Some selection criteria are first set up before the final team selection.
• Step 6: Pick up the top 10 players from the batting and bowling records. And
compare each player's statistics with other characteristics like partnership records, all-
rounder records, and wicketkeeper records.
• Step 7: After ranking each player, find the final rank of the best player's selection
through the linear expression of ranks.
• Step 8: Based on the final ranking selection out 11 players from 15 players’ bench.
• Step 9: Finally different rank priority along with final rank suggests which player
belongs to which characteristics.
• Step 10: According to the tournament style opponent's strength has been evaluated as
per high-rank team selection in a round-robin fashion.

Weight Chart:

12
6 Evaluation
The data analysis and findings determine the actual objectives of the research title. After
completion of data collection through web scraping and data pre-processing, one-by-one
datasets have been analysed for player evaluation through rank estimation.

6.1 Experiment- Important queries after rank estimation from the

datasets
(i) Batting Section:

a. Who made most 6s in the ODI?

13
In the figure above, it has been observed that after the player’s ranking in the batting section a
query about who has made the maximum of 6s compared to others was asked for verification
of the ranking dataset, and it showed the accurate result.

b. Who made most Runs in the ODI?

In the above screenshot, it has been observed that after the player’s ranking in the batting
section, a query about who has made a maximum of runs compared to others was asked for
verification of the ranking dataset.

(ii) Bowling Section:

a. Who have SR less than 30 but one of first 10 persons wicket takers?

In the above screenshot, it has been observed that after the player’s ranking in the bowling
section, a query about who has a strike rate of less than 30 but one of the first 10 persons
wicket takers with compared to others was asked for verification of the ranking dataset.

b. Who have maidens greater than 2 but one of first 10 persons wicket takers?

In the above screenshot, it has been observed that after the player’s ranking in the bowling
section, a query about who has maidens greater than 2 but one of the first 10 persons wicket
takers with compared to others was asked for verification of the ranking dataset.

14
 (iii) Fielding Section:
a. Who have most stumping among the wicket keepers?

In the above screenshot, it has been observed that after the player’s ranking in the fielding
section, a query about who has the most stumping among wicket keepers with compared to
others was asked for verification of the ranking dataset.

b. Who have caught most catches?

In the above screenshot, it has been observed that after the player’s ranking in the fielding
section, a query about who has caught the most catches with compared to others was asked
for verification of the ranking dataset.

(iv) All-rounder Section:

a. List the first 10 ranked Players who's bowling average greater than 30

In the above screenshot, it has been observed that after the player’s ranking in the all-rounder
section, a query about listing the first 10 ranked players who are bowling average greater than
30 compared to others was asked for verification of the ranking dataset.

b. List the first 10 ranked Players who's bowling average as well as batting
average greater than 60

In the above screenshot, it has been observed that after the player’s ranking in the all-rounder
section, a query about listing the first 10 ranked players whose bowling average as well as
batting average greater than 60 compared to others was asked for verification of the ranking
dataset.

15
 (v) Player-partnership Section:
a. Who has played 30 overs partnership among most top 10 partnership players

In the above screenshot, it has been observed that after the player’s ranking in the player-
partnership section, a query about who has played 30 overs partnership among most top 10
partnership players compared to others was asked for verification of the ranking dataset.

b. Who has played for top partnership against Australia

In the above screenshot, it has been observed that after the player’s ranking in the player-
partnership section, a query about who has played for top partnership against Australia
compared to others was asked for verification of the ranking dataset.

6.2 Win/loss ratio analysis from the datasets extracted

Against Australia Against West Indies

India won from 1st bat innings = 8 India won from 1st bat innings = 11
India lost from 1st bat innings = 7 India lost from 1st bat innings = 1
Table 2: India batting in 1st innings (Records from last 4 years)

16
 Against Australia Against West Indies

India won from 1st bowl innings = 8 India won from 1st bowl innings = 11
India lost from 1st bowl innings = 7 India lost from 1st bowl innings = 1
st
Table 3: India bowling in 1 innings (Records from last 4 years)

In table 2 it has been observed that when India took the first batting in the innings and had a
face against Australia and West Indies then the ratio against win to loss is measured.
Similarly, from taking bowling in the innings the ratio is also measured for observing
winning and losing against Australia and West Indies. It is secondary important to select a
team for doing up and down ranked selection. It is also preferred for selecting a new player
and a rising player from the high-ranked players. The result observed from the above graphs
is that while considering teams such as Australia and West Indies, the win ratio is huge when
team India is bowling in the first innings.

6.3 Add/remove player’s condition

In this figure above, SV Samson has a very low value of weight for rank partnership so this
player cannot be selected. So looking for a new player under acceptable and marginal
conditions like that good SR, average partnership, and good fielding rank also in the batting
side. After searching the players, the results have been taken out.

17
Three players (RA Jadeja, MS Dhoni, and KM Jadhav) are reviewed for team selection. Now
the question is:
• Which player is having best partnership amongst RA Jadeja, MS Dhoni, and
KM Jadhav?

From the above table, it is seen that RA Jadeja has a better partnership than MS Dhoni and
KM Jadhav. Hence, RA Jadeja is selected.

6.4 Final Rank Estimation

From the above table, it is observed that the final ranks of the selected 15 players are most
essential to select the 11 players instead of calculation of Combination = .
For instance, player V Kohli has the greatest batting as well as fielding rank. So, this player is
considered a batsman. After performing the calculation, and cumulating all the weights of the
respective players, the final squad has been selected.

6.5 Discussion
From the valuation part, it has been observed how the best players are selected through
the linear and weightage-based rank system. It has also been verified how the rank selection
is predicted from the match (India v Australia 2020 3rd ODI Sun 19 January, 13:30 Local
18
(13:30 IST) M.Chinnaswamy Stadium, Bangalore, India). In this match, India had batted at
first with a run rate of 6.08 and scored 289 with a fall of wickets 3. And Australia lost this
match they had scored 286 runs at a run rate of 5.72 and fell off nine wickets. Below is the
player’s selection for this match.

Actual Player Predicted Ranked Players Player Matched

(Who played the match)
Rohit Sharma V Kohli V Kohli
KL Rahul S Dhawan S Dhawan
Virat Kohli RG Sharma RG Sharma
Shreyas Iyer Kuldeep Yadav Kuldeep Yadav
Manish Pandey RA Jadeja Mohammed Shami
Shikar Dhawan HH Pandya JJ Bumrah
Ravindra Jadeja YS Chahal SS Iyer
Mohammad Shami Mohammed Shami RA Jadeja
Kuldeep Yadav SN Thakur KL Rahul
Navdeep Saini KL Rahul
Jasprit Bumrah JJ Bumrah
SS Iyer
B Kumar
RR Pant
Shubman Gill
Table 4: Comparison between Prediction and Matched Player
From this above table, it can be found that the total player matched is 9 out of 11 players who
have been selected to play the match. The accuracy rate is about 81.8%. The rest of the work
which is included in the report has been added to the configuration manual.

7 Conclusion and Future Work

In this report, to achieve the outcome of the research objective KDD methodology has
been applied. Initially, the data is extracted from the sports website called ‘espncricinfo’ with
the required query from the last 4 years to evaluate the model. The portal is based on
selection features like batting, bowling, all-rounder, fielding, wicket-keeping, player-
partnership, and team first-innings win ratio for both batting and bowling are web-scrapped
using python coding. Then the data quality has been assessed, all the null values are checked
and replaced with zeroes. Later, the data was transformed into proper datatypes for further
steps. After getting the data ready, a ranking-based indexing strategy based on linear
programming is applied to the attributes of each player on every selective feature. Each
player's features and other aspects are thoroughly analysed using the linear discriminate
approach to determine their ranking. After achieving the ranks for the attributes, the total
weightage is found and on the criteria of 9 batsmen, (including both all-rounders and wicket-
fielders) and 6 full-time bowlers are selected by considering all the aspects. After the final 15
are selected, for checking the accuracy it has been cross validated with one of the previous
matches and it is found that nine players have been matched which says the experiment has
an accuracy of 81.8%. In future work the classifiers can be used for finding the rank
estimation when new youngsters emerge onto the ground and based on the importance of that
19
match, players can be substituted and given rest for upcoming big tournaments. The accuracy
of the findings can be improved without compromising the factors that help in the selection.
8 Acknowledgement
I would like to thank my supervisor Dr. Cristina Hava Muntean for her constant support
and feedback without whom this project would be accomplished. I also want to thank the NCI
library mentor (Shannon Mallon) for the support without hesitation. And all the staff NCI
staff directly or indirectly help me reach this stage with all knowledge and guidance. And
finally, my lovable parents for standing by me at all odds and encouraging me.
References
Agrawal, P. and Ganesh, T., 2020, April. Selection of Indian Cricket Team in ODI using
Integer Optimization. In Journal of Physics: Conference Series (Vol. 1478, No. 1, p.
012001). IOP Publishing.

Bhattacharjee, D. and Saikia, H., 2016. An objective approach of balanced cricket team
selection using binary integer programming method. Opsearch, 53(2), pp.225-247.

Chakraborty, S., Kumar, V. and Ramakrishnan, K., 2018. Selection of the all-time best World
XI Test cricket team using the TOPSIS method. Decision Science Letters, 8(1), pp.95-108.

Hussain, A., Qiang, Y., Bilal, A.Q.M., Ullah, U. and Ullah, N., 2019. Region-based teams
ranking in the game of cricket using PageRank algorithm. Int. J. Comput. Appl, 177(16),
pp.10-15.

Hussain, A., Yan, Q., Bilal, M.A.Q., Wu, K., Zhao, Z. and Ahmed, B., 2019. Region-wise
ranking for one-day international (ODI) cricket teams. International Journal of Advanced
Computer Science and Applications, 10(10).

Jayanth, S.B., Anthony, A., Abhilasha, G., Shaik, N. and Srinivasa, G., 2018. A team
recommendation system and outcome prediction for the game of cricket. Journal of Sports
Analytics, 4(4), pp.263-273.

Jha, A., Kar, A.K. and Gupta, A., 2022. Optimization of team selection in fantasy cricket: a
hybrid approach using recursive feature elimination and genetic algorithm. Annals of
Operations Research, pp.1-29.

Kissell, R. and Poserina, J., 2017. Optimal sports math, statistics, and fantasy. Academic
Press.

Mahbub, M.K., Miah, M.A.M., Islam, S.M.S., Sorna, S., Hossain, S. and Biswas, M., 2021,
November. Best Eleven Forecast for Bangladesh Cricket Team with Machine Learning
Techniques. In 2021 5th International Conference on Electrical Engineering and Information
& Communication Technology (ICEEICT) (pp. 1-6). IEEE.

Perera, W.A.S.C., 2021. Predicting an Optimal Sri Lankan Cricket Team for ODI Matches
According to the Nature of the Game (Doctoral dissertation).

20
Prakash, C.D., Patvardhan, C. and Singh, S., 2016. A new machine learning based deep
performance index for ranking IPL T20 cricketers. International Journal of Computer
Applications, 137(10), pp.42-49.

Premkumar, P., Chakrabarty, J.B. and Chowdhury, S., 2020. Key performance indicators for
factor score-based ranking in One Day International cricket. IIMB Management
Review, 32(1), pp.85-95.

Shah, P. and Patel, M.N., 2018. Ranking the cricket captains using principal component
analysis. Int. J. Physiol. Nutr. Phys. Educ, 3(2), pp.477-483.

Shah, P., 2017. New performance measure in Cricket. ISOR Journal of Sports and Physical
Education, 4(3), pp.28-30.

Singla, S. and Shukla, S.S., 2020. Integer optimisation for Dream 11 cricket team
selection. International Journal of Computer Sciences and Engineering.

Vetukuri, V.S., Sethi, N. and Rajender, R., 2021. Generic model for automated player
selection for cricket teams using recurrent neural networks. Evolutionary Intelligence, 14(2),
pp.971-978.

Viva Question and answers

Q1: How many players were there in total in the dataset? Was there a way to automate the
ranking/team creation with machine/deep learning solutions?
Ans1: In this research project, there are total 5 datasets based on the feature selection such as
players based on their batting performance, bowling performance, all-rounder performance,
fielding impact, and player-partnership built. In here, all datasets are stored in solution
artifact folder which has already been submitted in the moodle. The batting data has 50
players named as ‘bat.csv’, bowling dataset has 50 players named as ‘bowl.csv’, all-rounders
dataset has 50 players named as ‘allround.csv’, fielding dataset has 50 players named as
‘fielding.csv’, and player-partnership dataset has 446 rows with each player batting with
other players in that time frame named as ‘partnership.csv’.
There can be many machine/deep learning methods to arrive at the solution of ranking
or team creation. In this scenario, rank method was used and performed by linear
discrimination method (which is one of the machine learning approach) and weightage
calculation based.

Q2: Are there other queries that could have been used in the ranking process?
Ans2: Other than the feature selection that has been mentioned in the introduction of the
report, wicket-keeping is also one of the important factors that need to be considered while
ranking the players. Apart from these in the historical data used in the project, there can be
filtered as players who can play best on different types of pitches/grounds. One more query
that can be used is the weather condition on the day they are playing, it can be a hot climate, a
cold place, a day match, or a night match. These could have been used if a proper dataset is
found. If the queries have to be within the dataset used, then the query can be the dependency
of the bowling pair that can be used to break the partnership of the opposite team or the
bowling pair that can be used in powerplay or death over bowling. For making rank-based
selection there are queries for best feature selection.

21
Q3: Does the presence of a player in the team influence the performance of another
player? (e.g. when player A plays, player B plays better, so it is good to have player A and B
together)
Ans: Yes, the presence of one player will definitely influence the performance of another
player. In this case, one of the feature selections which has been included is player-
partnership. Here, there are situations where the batsman has been performing better when
another batsman whom he is comfortable with is playing on the other end of the batting.
While ranking the player partnership, queries have been used to check if were there any
influencing players that need to be added to the final list.

Q4: Would you be able to predict a player's performance based on the historical data? How
would you do that?
Ans: Yes, it is possible to predict the player’s performance based on historical records. In this
situation, the first datasets were gathered through web scrapping from www.espncricketinfo,
which are completely based on their statistics in past years. For instance, the
performances/statistics of the players based on feature selections were taken from the last
four years. And then take the data of the players and by linear discriminatory method and
weightage calculations of each attribute the players' statistics are checked and the rank of
each player as well as the player's performance can be achieved. Finally, the queries are used
to check the accuracy of the values.
Q5: Briefly present the limitations of your research work.
Ans: When it comes to limitations, there are a few aspects that are not considered while
predicting the players. In this work, only players who have played for India have been
considered. Youngsters who have played well but didn’t get the chance to play for the Indian
team are not included. Another limitation is that this is the method applied on Indian one-day
international, but when it comes to the 20-over format the players need to be more
aggressive, runs should be scored quickly, and the bowlers should focus more on taking
wickets than reducing the runs. Some of the conditions are the player's health condition
before, during, and after the match is not considered here. So, one cannot predict the player's
fitness without these. Also, in this project players' emotion is not considered, and their mental
being was not included. Hence in the project, further research can be carried out through
these parameters to obtain well suit players for match specific.

22