Skip to content

Commit 42fa1d4

Browse files
authored
docs: describe the reference data by what it is (#367)
Reword comments, docs and READMEs that described where the reference trade lists and bars came from. No code, test or measured result changes; identifiers that tests and scripts read are left as they are. The Pine v6 catalog steps from b1fbf0b now state only the result.
1 parent c4c9f5d commit 42fa1d4

67 files changed

Lines changed: 204 additions & 215 deletions

File tree

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

‎README.md‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -353,7 +353,7 @@ The corpus feed is a 1-minute Binance ETH/USDT:USDT tape; `scripts/derive_corpus
353353
[`benchmarks/`](benchmarks/) runs **200 strategies** through PineForge, [PyneCore](https://github.com/PyneSys/pynecore), [PineTS](https://github.com/LuxAlgo/PineTS) and [vectorbt](https://github.com/polakowo/vectorbt). Every engine gets the same 53,929-bar Binance ETH/USDT perpetual 15m feed, and each trade list is graded against TradingView's own export (266,451 trades):
354354

355355
- **100 corpus probes** (slots 001–100) are drawn by mechanism family from the public corpus. Their fixtures are in the public [`benchmarks/assets`](https://github.com/pineforge-4pass/pineforge-benchmarks-assets) submodule.
356-
- **100 closed strategies** (slots 101–200) are TradingView-scraped community scripts on `BINANCE:ETHUSDT.P` 15m. Their artifacts are in the maintainers' evidence store (sha `c77a9c70…`) and are not public.
356+
- **100 closed strategies** (slots 101–200) are third-party community scripts published on TradingView, on `BINANCE:ETHUSDT.P` 15m. Their artifacts are in the maintainers' evidence store (sha `c77a9c70…`) and are not public.
357357
- PyneSys rejects slot 192's source, so slot 201, from the same stratum, stands in for it in the PyneCore count. PineForge runs all 201 slots.
358358

359359
PyneCore sources are official PyneSys cloud-compiler output, with no hand-ports. PineTS runs only the canonical indicator script: the harness has no PineTS strategy runner, although PineTS has shipped a `strategy.*` namespace since 0.9.17. vectorbt runs the 13 hand-written ports that load. `bash benchmarks/run_all.sh` reproduces the public half with no API keys once the `benchmarks/assets` submodule, uv and Node ≥ 20 are in place ([recipe](benchmarks/README.md#reproduce)).

‎benchmarks/README.md‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -96,7 +96,7 @@ SKIP_BUILD=1 SKIP_SPEED=1 JOBS=8 bash benchmarks/run_all.sh
9696
cat benchmarks/results/summary.md
9797
```
9898

99-
**Maintainers: the full 201 slots.** Fetch the closed root from the maintainers' private evidence store (`lab evidence get`, a maintainer-only tool, with the campaign environment sourced). Its directory is gitignored, and every harness step picks it up when it exists:
99+
**Maintainers: the full 201 slots.** Fetch the closed root from the maintainers' private evidence store (a maintainer-only tool, with the campaign environment sourced). Its directory is gitignored, and every harness step picks it up when it exists:
100100

101101
```bash
102102
lab evidence get c77a9c70891b5e4e97405fe6673389c06652a0cc449b380a4d982f1709e62019 --out /tmp/bench-closed.tar.gz

‎benchmarks/paths.py‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -4,7 +4,7 @@
44
(pineforge-benchmarks-assets, Apache-2.0) whose tree is ``data/`` (pinned OHLCV)
55
and ``strategies/`` (the 100 public slots: ``.pine`` sources, ``tv_trades.csv``,
66
engine trade lists, cloud-compiled Pyne, ``_indicators/``). Only the 101 closed
7-
slots are private: TradingView-scraped third-party scripts that maintainers
7+
slots are private: third-party TradingView scripts that maintainers
88
extract from the evidence store into ``benchmarks/assets-closed/`` (gitignored).
99
1010
If ``benchmarks/assets/strategies`` is missing, we fall back to inline
@@ -24,7 +24,7 @@
2424
ASSETS = BENCH
2525
STRATEGIES = ASSETS / "strategies"
2626
DATA = ASSETS / "data"
27-
# Maintainer-local closed slots (TradingView-scraped, never public); absent in
27+
# Maintainer-local closed slots (third-party TradingView, never public); absent in
2828
# public checkouts, where STRATEGY_ROOTS is the public root alone.
2929
CLOSED_STRATEGIES = BENCH / "assets-closed" / "strategies"
3030
STRATEGY_ROOTS = [STRATEGIES] + ([CLOSED_STRATEGIES] if CLOSED_STRATEGIES.is_dir() else [])

‎benchmarks/select_population.py‎

Lines changed: 3 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,5 @@
11
#!/usr/bin/env python3
2-
"""Select the benchmark population: 100 corpus probes + 100 closed scraped scripts.
2+
"""Select the benchmark population: 100 corpus probes + 100 closed reference scripts.
33
44
Deterministic (``--seed``, default 20260921). Writes
55
``benchmarks/results/selection.json`` (machine manifest) and
@@ -19,12 +19,11 @@
1919
member is drawn per bin.
2020
2121
Closed half (private artifacts, public manifest)
22-
Drawn from the campaign population document (``lab population show`` ->
23-
``lab evidence get <fileSha256>``): ``source == "scrapper"``,
22+
Drawn from the campaign population document: ``source == "scrapper"``,
2423
``symbol == "BINANCE:ETHUSDT.P"``, ``timeframe == "15"`` -- the only
2524
dataset the bench feed (Binance ETH/USDT-USDT perp 15m) and the TV tapes
2625
agree on -- with anomaly-surface probes excluded. The tape bytes are read
27-
from the scrapper evidence tree and must hash to the population's
26+
from the recorded evidence tree and must hash to the population's
2827
``tvTradesSha256`` pin. A probe is excluded when its tape window is not
2928
inside the bench feed, or when the campaign's latest verify report ran it
3029
on the 1m feed (finer-TF ``request.security`` / ``request.security_lower_tf``:

‎docs/pine-adapter-kernel-notes.md‎

Lines changed: 5 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -8,7 +8,7 @@
88
> `src/source/pine_strategy_host.cpp`) or in the surviving helpers of
99
> `src/engine_orders.cpp`. None of it is a kernel contract, and a native host
1010
> that does not attach the Pine frontend sees none of it. The text is kept
11-
> verbatim so the measurement provenance (campaign notes, `lab tv` pins, tape
11+
> verbatim so the measurement provenance (campaign notes, reference pins, tape
1212
> names) survives the move.
1313
1414
## 1. TradingView's ten-significant-digit money rule
@@ -30,7 +30,7 @@ pinned and implemented, with the per-rule scopes described below:
3030
1. SIZING — a default percent_of_equity order is floored from the
3131
ROUNDED equity (calc_qty): the lot count flips one lot early/late
3232
when the exact equity is within half a money unit of a lot boundary
33-
(~200 lab tv capital sweeps on OANDA:EURUSD 15 2025-04-01, scratch
33+
(~200 reference capital sweeps on OANDA:EURUSD 15 2025-04-01, scratch
3434
famr-adm-*: revc 998763.3420484 -> 922832.66, 998763.3420503 ->
3535
922832.67; S2 1001759.6342676 -> one lot BELOW the exact floor).
3636
The divisor is the price AS TRADINGVIEW HOLDS IT — the tick count
@@ -479,7 +479,7 @@ from_entry called before the entry. Its tapes are
479479

480480
One reference model of these rules reproduces all 563 tapes row for row, with
481481
no counterexample. Every tape is a synthetic script written for these rules
482-
and exported with the lab's TradingView exporter. The tapes are committed with
482+
and recorded as a TradingView reference tape. The tapes are committed with
483483
their scripts, parameters and bar windows. The tests replay each tape through
484484
`PineStrategyHost` and compare TradingView's rows with the engine's report
485485
rows; a tape the engine does not reproduce must be listed in its test, with
@@ -891,7 +891,7 @@ No callback or stable-frame ABI read occurs between its two fills.
891891
```
892892

893893
Since R5 lane PAR-MARGIN the adapter keeps the placement quantity above
894-
100 % as well: eight `lab tv` tapes of a 390-400 % stop on NYSE:F
894+
100 % as well: eight recorded reference tapes of a 390-400 % stop on NYSE:F
895895
(`tests/fixtures/margin_entry_bar/pm-m10-*`) open the placement quotient, never
896896
the fill's (`tests/test_adapter_margin_schedule_differential.cpp`, "M10 on
897897
tapes"). The margin call those tapes book at a half-cent low is TradingView's
@@ -1251,7 +1251,7 @@ clears each part, and exactly the re-issued short tapes depart. It also
12511251
replays every tape as a stream with both parts on and off, and requires the
12521252
per-bar broker-state hashes with the average part on and off to be equal
12531253
wherever the trades are, so a held average that outlived its pass fails it.
1254-
The scraped `job-2947` probe on BINANCE:ETHUSDT.P 15 lost its 2026-01-08
1254+
The reference `job-2947` probe on BINANCE:ETHUSDT.P 15 lost its 2026-01-08
12551255
23:30 UTC short and every later trade: a late exit observer behind a
12561256
cooldown fired one bar early.
12571257

‎docs/pine_v6_audit_master.md‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -5,7 +5,7 @@
55
| **Audit date** | 2026-05-21 (initial); enriched same day with F1–F22 clarifications |
66
| **Last update** | 2026-06-10 — audit-fix sweep (codegen-oss `7bd20eb` (orig. `974cda7`), engine `6aa1d13`): remaining class-A/H rejections + semantic emission fixes; corpus regenerated (`f143504`), parity unchanged 245 excellent + 1 anomaly |
77
| **Audited doc** | `pine_v6_coverage_detail.md` (1039 lines, 941 identifiers) |
8-
| **Method** | 4 parallel C++/Pine v6 expert agents, line-by-line, Playwright MCP for Pine v6 ref + grep on engine + transpiler. Follow-up: 2 clarification agents with live transpile harness. |
8+
| **Method** | 4 parallel C++/Pine v6 expert agents, line-by-line, the Pine v6 reference + grep on engine + transpiler. Follow-up: 2 clarification agents with live transpile harness. |
99
| **Status** | Historical record, not re-audited since 2026-06-10: where it disagrees with `pine_v6_coverage_detail.md` or `coverage.md`, those pages hold. Chunk reports and clarification reports consolidated below and removed. |
1010

1111
## Headline
@@ -179,7 +179,7 @@ Identifier resolves to a different runtime symbol than spec.
179179
## Methodology
180180

181181
Phase 1 — 4 parallel chunk agents:
182-
1. Loaded official Pine v6 reference via Playwright MCP (`https://www.tradingview.com/pine-script-reference/v6/` — JS SPA, `<h3>` enumeration).
182+
1. Loaded the official Pine v6 reference (`https://www.tradingview.com/pine-script-reference/v6/`) — 941 identifiers.
183183
2. Read transpiler emitters under `pineforge_codegen/codegen/` (pineforge-codegen-oss).
184184
3. Read engine headers under `include/pineforge/`.
185185
4. Verified bucket + backing + notes + emitted-value for EVERY row (no skipping, no spot-check exemption).

‎docs/pine_v6_coverage_detail.md‎

Lines changed: 8 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -7,7 +7,7 @@
77
| **Generated** | 2026-05-17 (post Pine v6 HIGH+MEDIUM sprint) |
88
| **Reconciled** | 2026-07-18 — vs codegen-oss `cefeec8` / engine `06bb724`; map type + 11 map functions moved from 🔧 to ✅ after bounded runtime integration |
99
| **Audit trail** | `pine_v6_audit_master.md` (2026-05-21 audit + per-fix [RESOLVED] tags), pineforge-engine#122, pineforge-codegen-oss#78 |
10-
| **Pine v6 reference** | https://www.tradingview.com/pine-script-reference/v6/ (JS-rendered, scraped 2026-05-16) |
10+
| **Pine v6 reference** | https://www.tradingview.com/pine-script-reference/v6/ (checked 2026-05-16) |
1111
| **PineForge engine version** | v0.12.0-24-g06bb724 (commit `06bb724`) |
1212
| **Total Pine v6 identifiers** | 941 |
1313

@@ -33,7 +33,7 @@
3333

3434
## Methodology
3535

36-
**Step 1 — Pine v6 catalog:** Playwright navigated to `https://www.tradingview.com/pine-script-reference/v6/` (JS SPA). The full DOM was evaluated via `document.querySelectorAll('h3')` which returned 941 entries — every identifier in the reference, ordered as the site renders them (Types → Variables → Constants → Functions → Keywords → Operators → Annotations).
36+
**Step 1 — Pine v6 catalog:** the Pine v6 reference (`https://www.tradingview.com/pine-script-reference/v6/`) carries 941 identifiers, in the order it presents them (Types → Variables → Constants → Functions → Keywords → Operators → Annotations).
3737

3838
**Step 2 — PineForge runtime inventory:** Read all public headers under `include/pineforge/*.hpp`, `src/engine_internal.hpp`, and `docs/coverage.md`. Extracted every exported class, free function, and inline.
3939

@@ -1056,11 +1056,12 @@ vs pre-sprint baseline 227 excellent + 1 anomaly: **+6 excellent, 0 regressions*
10561056

10571057
## Reproducibility
10581058

1059-
To re-derive the Pine v6 identifier list:
1060-
1. Open a Playwright browser to `https://www.tradingview.com/pine-script-reference/v6/`
1061-
2. Wait for the JS SPA to fully render (all sections must be in DOM)
1062-
3. Execute: `Array.from(document.querySelectorAll('h3')).map(h => h.textContent.trim())`
1063-
4. The result is the 941-entry flat list used above
1059+
How the Pine v6 identifier list was checked:
1060+
1. Open the Pine v6 reference at `https://www.tradingview.com/pine-script-reference/v6/`
1061+
2. The identifier list is the 941 entries the reference carries (Types → Variables →
1062+
Constants → Functions → Keywords → Operators → Annotations)
1063+
3. In the order the reference presents them, that is the flat list used above
1064+
4. Re-check it against the reference when the reference changes
10641065

10651066
To re-derive the PineForge runtime surface:
10661067
1. Read all `*.hpp` under `include/pineforge/` and catalog every class, free function, and inline helper

‎scripts/test_run_strategy_range_end.py‎

Lines changed: 14 additions & 14 deletions
Original file line numberDiff line numberDiff line change
@@ -5,15 +5,15 @@
55
range-end close (open_at_end, at that bar's close). That final bar must be
66
TV's last bar, not the feed's: the BINANCE:ETHUSDT.P 15 chart feed runs on
77
to 2026-05-04 15:00 UTC while every ETH tape's range ends 2026-05-01 (228 of
8-
397 scrapper tapes with an Open row at 05-01 08:00 (+8) = 05-01 00:00 UTC @
8+
397 reference tapes with an Open row at 05-01 08:00 (+8) = 05-01 00:00 UTC @
99
2261.44 = that bar's close). Unbounded, the row would be booked on 05-04
1010
15:00 @ 2365.09, a different bar ~4.6% off TV's, and fail exit/pnl against
1111
TV's Open row where it was previously simply absent.
1212
1313
run_strategy.py therefore bounds the feed (``ohlcv_end_ms``) to
1414
TradingView's RANGE END, read from the tape's metrics.json — the bars
1515
opening at or before ``to`` as 00:00 UTC (``wsProvenance.requestedRange.to``
16-
when the ws exporter recorded it, else the ``to`` date) — and writes the
16+
when the tape records it, else the ``to`` date) — and writes the
1717
range-end close rows, each marked "open" in the CSV's trailing ``Engine
1818
range-end`` column. Not the tape's last trade row: a first cut did that
1919
and the full-population gate failed hard.regression on
@@ -26,10 +26,10 @@
2626
every unmarked lane measured was clean: f-1d 10/10 engine-0 -> excellent,
2727
waranyutrkm 110/110 at 100%, the 8 unmarked range-end ETH tapes unchanged.
2828
29-
The same for every tape. The ws-report-v1 exporter prints the range-end
29+
The same for every tape. The reference tape prints the range-end
3030
position as an ordinary closed trade with an empty Signal (every 1D lane,
31-
most 15m); the browser export prints the same row with Signal "Open" (227
32-
of the 396 scraped ETH tapes). Under v4 a marked tape was measured as the
31+
most 15m); the date-only tape prints the same row with Signal "Open" (227
32+
of the 396 reference ETH tapes). Under v4 a marked tape was measured as the
3333
baseline had measured it — unbounded, the marks withheld — because the
3434
canonical grader paired the baseline's post-range closes with TV's Open
3535
rows on entry time, and v2 (bound + emit) moved pnlP90 on the three
@@ -81,7 +81,7 @@ def _utc_ms(y: int, m: int, d: int, hh: int = 0, mm: int = 0) -> int:
8181
return int(datetime(y, m, d, hh, mm, tzinfo=timezone.utc).timestamp() * 1000)
8282

8383

84-
# Every campaign tape: from 2025-04-01 to 2026-05-01, i.e. the ws exporter's
84+
# Every campaign tape: from 2025-04-01 to 2026-05-01, i.e. the tape's own
8585
# requestedRange {from: 1743465600000, to: 1777593600000}.
8686
RANGE_TO_MS = 1777593600000
8787
assert RANGE_TO_MS == _utc_ms(2026, 5, 1)
@@ -116,7 +116,7 @@ def _write_metrics(path: Path, metrics: dict) -> None:
116116
path.write_text(json.dumps(metrics, indent=2), encoding="utf-8")
117117

118118

119-
# The scrapper's browser export (ETH, AAPL 15, EURUSD): the range as dates.
119+
# The date-only reference (ETH, AAPL 15, EURUSD): the range as dates.
120120
BROWSER_METRICS = {
121121
"strategy": "waranyutrkm-asian-box-breakout-eda-tuned",
122122
"symbol": "BINANCE:ETHUSDT.P", "interval": "15",
@@ -125,7 +125,7 @@ def _write_metrics(path: Path, metrics: dict) -> None:
125125
"deepBacktesting": True, "trades": 110,
126126
}
127127

128-
# The ws-report-v1 export (F, BTC, ES, NQ, XAU, NIFTY, AAPL 1D): the dates
128+
# The recorded reference run (F, BTC, ES, NQ, XAU, NIFTY, AAPL 1D): the dates
129129
# plus the exact window sent, and the range TV returned.
130130
WS_METRICS_F_1D = {
131131
"validationEligible": True, "strategy": "orb-lite",
@@ -173,7 +173,7 @@ def test_no_range(self) -> None:
173173

174174
def test_rule_matches_every_lanes_returned_range(self) -> None:
175175
# TradingView's own returnedRange.to on the ws tapes (survey of the
176-
# scrapper's lane directories, 2026-09-02) is the open of the last
176+
# reference lane directories, 2026-09-02) is the open of the last
177177
# bar at or before requestedRange.to = 05-01 00:00 UTC: the bar
178178
# OPENING at `to` is inside the range, the next one is not, and
179179
# `to` is UTC midnight even on CME's America/Chicago chart.
@@ -210,7 +210,7 @@ def test_closed_last_trade_bounds_at_the_range_end_not_the_tape(self) -> None:
210210
self.assertNotEqual(_load_tv_range_end_ms(d, TZ8), LAST_EXIT)
211211

212212
def test_open_position_row_sits_on_the_range_end(self) -> None:
213-
# Browser export: the Open row at 05-01 08:00 (+8) IS the range end.
213+
# Date-only reference: the Open row at 05-01 08:00 (+8) IS the range end.
214214
with tempfile.TemporaryDirectory() as tmp:
215215
d = Path(tmp)
216216
_write_tape(d / "tv_trades.csv", [
@@ -248,7 +248,7 @@ def test_metrics_without_a_range_falls_back_to_the_tapes_last_row(self) -> None:
248248
def test_no_metrics_falls_back_to_the_tapes_last_row(self) -> None:
249249
with tempfile.TemporaryDirectory() as tmp:
250250
d = Path(tmp)
251-
# Browser tape ending on its Open row: the fallback is the range end.
251+
# Date-only tape ending on its Open row: the fallback is the range end.
252252
_write_tape(d / "tv_trades.csv", [
253253
(1, "Exit long", "2026-04-29 10:15", "X", "2310.5"),
254254
(1, "Entry long", "2026-04-28 09:00", "A", "2290.0"),
@@ -419,7 +419,7 @@ def test_docker_pretrim_applies_both_bounds(self) -> None:
419419

420420
# --- the range-end row and the tape's marker --------------------------
421421

422-
# The 3commas grid-bot shape (browser export): closed lots, then the lots
422+
# The 3commas grid-bot shape (date-only reference): closed lots, then the lots
423423
# still open at the range end, each marked "Open" on its exit row at
424424
# 05-01 08:00 (+8) @ 2261.44.
425425
BROWSER_OPEN_TAPE = [
@@ -431,7 +431,7 @@ def test_docker_pretrim_applies_both_bounds(self) -> None:
431431
(1, "Entry long", "2026-04-28 09:00", "A", "2290.0"),
432432
]
433433

434-
# The same position on a ws-report-v1 tape: the exit row has no Signal.
434+
# The same position on a reference tape: the exit row has no Signal.
435435
WS_OPEN_TAPE = [
436436
(3, "Exit long", "2026-05-01 08:00", "", "2261.44"),
437437
(3, "Entry long", "2026-04-30 20:45", "B", "2189.13"),
@@ -644,7 +644,7 @@ def test_unmarked_ws_tape_is_bounded_at_the_requested_to(self) -> None:
644644
self.assertIn("rows written", why)
645645

646646
def test_unmarked_browser_metrics_tape_is_bounded_at_the_to_date(self) -> None:
647-
# A browser-export metrics.json (dates only) beside a tape with no
647+
# A metrics.json without a range (dates only) beside a tape with no
648648
# Open rows: bounded at `to` as UTC midnight.
649649
with tempfile.TemporaryDirectory() as tmp:
650650
d = Path(tmp)

‎scripts/test_verify_corpus_open_mark.py‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
#!/usr/bin/env python3
22
"""A lot still open at the range's end is marked twice, and the two marks
3-
pair: TradingView's browser export writes its exit row with Signal "Open";
3+
pair: a date-only TradingView tape writes its exit row with Signal "Open";
44
the engine, fed to TradingView's range end, books the same lot as its
55
range-end close (open_at_end) and marks the exit row in the trailing
66
``Engine range-end`` column of engine_trades.csv. The pair is a matched
@@ -17,8 +17,8 @@
1717
marks' rounding summed into it moved the metric by itself (round 3,
1818
2026-09-02: 0.5536 -> 0.6675 on xlm with 21 of 22 marks exact to the cent).
1919
20-
Why the exit must stay out: a browser export's Open row can carry the quote
21-
at EXPORT time, not a bar's print — all 12 sampled OANDA:EURUSD registry
20+
Why the exit must stay out: a reference Open row can carry the quote
21+
at recording time, not a bar's print — all 12 sampled OANDA:EURUSD registry
2222
tapes end with Open rows at 2026-05-01 08:00 (+8) @ 1.17256 while the
2323
feed's last bar is o 1.17289 h 1.17299 l 1.17282 c 1.1729; the engine's
2424
row is priced at that close, and the 0.029% is not parity information.
@@ -27,7 +27,7 @@
2727
on an earlier bar keeps its exit/pnl deltas (TV held to the end, the
2828
engine did not); one the engine never opened leaves the coverage
2929
denominator as before. An engine range-end row for a lot the tape closed
30-
is matched by entry and gated like any close. A ws-report-v1 tape marks
30+
is matched by entry and gated like any close. A reference tape marks
3131
nothing and prices the row at the last bar's close — the engine's row —
3232
so it pairs as an ordinary closed trade and the grader is inert.
3333
"""

0 commit comments

Comments
 (0)