backtest: --record/--record-bars Parquet recorder + 2-year M5/M15 replay with indicator weights and calendar

Adds an opt-in recorder to the backtester so a replay keeps the evidence each
trade setup was decided on: every indicator's vote, value, confidence, base
weight, regime multiplier, effective weight and contribution on the primary and
both confirmation timeframes, penalties, the setup and outcome, and the Forex
Factory calendar events around the trade (4 h before entry to exit, all-day rows
on the New-York date). --record-bars adds one float32 row per analyzed bar with
the same votes plus 1 h/4 h/24 h forward moves in ATRs. Output is Parquet in a
folder per run with a manifest carrying the engine/gate fingerprints, the regime
weight table in force and the book MD5.

Nothing in the signal core changes: the trade book is byte-identical with
recording on or off (control-vs-recorded MD5 equal on both 2-year runs, M5
781 / +146.07R / PF 1.52, M15 793 / +117.62R / PF 1.41 at $0.65). Module 33
adds 12 tests (498 total): snapshot reproduces the engine's composite, absence
is no vote, calendar join edges/phases/holiday, write/load round-trip,
_replay_span parity with recording on, CLI flags, import and source guards.

Also: backtest/replay_scorecard.py (first-look scorecards, calendar buckets,
in-book and per-bar what-ifs, --check invariants), scripts/ff_calendar_scraper.py
restored from git history (the only way to regenerate the gitignored calendar
file), report in backtest/research/replay_recording_2026-09-09.md.

Claude-Session: https://claude.ai/code/session_01Nc9Ninzyi18QkWezrxeBvB

CLAUDE.md

@@ -12,7 +12,7 @@ Every numeric threshold in `config.py` and `pair_signals/xauusd.py` was backtest
 
 ```bash
 python test_all.py                      # QA suite, no network (mocked); prints its own count
-                                        #  (483 tests / 32 modules as of 2026-09-08)
+                                        #  (498 tests / 33 modules as of 2026-09-09)
 python main.py --pair XAUUSD --timeframe H1 --detail   # CLI signal report (--json for full dump,
                                         #  --panel for the advisory LLM agent panel, --outlook for the
                                         #  hourly AI market brief; the violent-move watch line appears
@@ -33,6 +33,10 @@ python backtest/gpu_backtest.py --pair XAUUSD --start 2025-01-01 --end 2025-06-0
 #            --export-features (append the 65 decision-time f_* columns; default export stays 63 cols)
 #            --fixed-rr-tp N (research-only: hard TP at N×risk, trail off; verdict
 #            2026-08-05: NOT shipped — backtest/research/fixedtp_2026-08-05.md)
+#            --record DIR [--record-bars] (research: manifest + trades/indicators/events parquet for
+#            every accepted trade, and one row per analyzed bar; the book is unchanged — pair every
+#            recorded run with a control run and show the MD5s equal; read with
+#            backtest.replay_record.load_run, first look with backtest/replay_scorecard.py DIR --check)
 ```
 
 **`test_all.py` is NOT pytest** — it uses a custom `@test("name")` decorator that runs each test immediately at import; don't grep for `def test_` or try `pytest test_all.py`. Run the whole file; it prints per-module results and exits non-zero on failure. There is no way to run a single test short of commenting modules out. Tests must stay green; every new gate/knob needs tests. Signal-log fixtures: use a WEEKDAY signal bar when a test exercises candle continuity — the outcome model treats weekend and 21:00 UTC gaps as scheduled, so Saturday-dated bars never produce a gap. `_candles_json` stores `df.tail(14)`, so a test snapshot longer than 14 bars silently drops its head. When patching weekend logic in tests, patch `providers.base._is_forex_weekend` — but know there are THREE independent signal-path weekend implementations (`providers/base.py`, `signal_aggregator.py`, and an inline copy in `backtest/gpu_backtest.py:_is_xau_blocked_by_calendar`) plus the advisory-only continuity copy in `advisory_outcome.py`; patching one does not reach the others.
@@ -108,3 +112,4 @@ Signal currency everywhere is `SignalStrength` (−2..+2). DataFrames are UTC tz
 - 2026-09-08 delivery-side fixes from the 3-day signal-log + telegram-log review (nothing in the signal core changed; 468 tests): (1) `_drop_stale_spread_gate_vetoes` (after every setup archetype, so the XAU_PB path is covered too) moves a core generator's stale `SPREAD GATE:` veto to `cleaned_warnings` when a later generator ships a setup — the 3-yr H1 and M15 books were regenerated before/after and are identical on every trade column (H1 253/+70.02R/PF 1.98, M15 1092/+151.90R/PF 1.38); only the exported `warnings` text differs, on 50/253 H1 and 358/1092 M15 trades, i.e. a fifth to a third of all A-grade alerts had been carrying "costs would eat the edge" beside a valid target; (2) Telegram alerts now require a trade setup (`_is_alert_candidate`, scan + breach paths; no-setup candidates log `alert skipped: … (gate reason)`); (3) every delivered alert / breach alert / violent-move heads-up logs an `… sent:` INFO line with the signal-log row id (the review had to reconstruct sends from panel rows); (4) `telegram.log` rotates daily × 30 instead of 3-hourly × 240 (old `_HH` files on the live box need one manual delete). Review facts (not evidence — three days): 8 setups, 1 right; strong readings (|score| ≥ 20) were followed by reversals in 16 of 18 episodes; hot watch level 0/25 bursts on M5 (calendar trigger never fired, 15 of 33 hot stamps were the 12–14 UTC slot on Labor Day — no holiday awareness anywhere); Oanda published no XAU bars 18:30–22:00 UTC on Labor Day and the bot scanned the stale bar 43×; the retry ladder recovered all 25 post-deploy 401 bursts (41 of 58 incidents sit on the :00 boundary); the live dashboard publisher ran only 01:00–02:20 UTC Sep 7 and is dead again.
 - 2026-09-06 stale-code cleanup (behaviour-neutral; suite green, M5/H1 12-month books byte-identical after every signal-side commit): `main.py --timeframe` now accepts M5/M15/H1/H4 and takes its confirmation TFs from the pair factory (the CLI's private table had paired M5 with M1+M15); `PlatformConfig.pip_value` is gone (`PAIR_INFO` is the source); removed as dead or superseded: `_get_affected_pairs`/`event_impact_on_pairs`, `SignalLogger.advance_with_bars`/`path`/`news_for`, `reset_llm_client`, the Telegram scan-summary/keyboard helpers and duplicate confirmation map, the 2026-07 Oanda-vs-TD research pager+study, `burst_profile.py`, `hour_analysis.py`, `scripts/compare_oanda_tiingo.py`, the never-implemented LLM-provider plan and the merged config.json spec. Second pass the same day (owner directive: keep only what the system needs): every research script and data file under `backtest/research/` (book_report, book_diff, the burst tools, sizing_model, commit_ladder, llm_news_eval + golden CSV, ff_calendar_scraper, calendar_impact, watch_rule_eval, the Dukascopy spread study + CSVs, the panel dry-test JSONL), the two completed study scripts in `scripts/`, the architecture diagram + generator and the data-source test plan were deleted; the campaign `.md` reports stay as the decision ledger. Git history before `953c5ad` has all of it. Third pass (owner: "remove dead functions as well"): the write-only pb/cp rejection counters, the unreachable `_TP_QUALITY_SCALE` B/C entries, the engine's unused DataFrame argument and M1 arm, the providers' generic-symbol guesses (unknown symbols now raise), the sub-$10 price format, the `=X` pair parsing and the unread `minutes_until`/`is_high_impact` news.json keys are gone.
 - 2026-09-08 news archive shipped (`news_archive.py`, 483 tests; nothing in the signal core changed, both news gating flags stay `False`, books untouched). Motivation: headlines only ever survived as a side effect of signal logging and `signals.db` was rotated around 2026-09-04, destroying the 3-week / 4,054-headline corpus the news study was written on — with `logs/` gitignored, git could not recover it and nobody noticed for four days. Day-one corpus after backfilling both DBs on disk: **1,016 distinct headlines, 846 of them DeepSeek-labelled**, growing ~200/day (~1,400/week). The point of collecting it: distil DeepSeek's headline scoring (78/78 on the 2026-08-06 golden set) into a numpy-only local model so the fallback stops being the keyword scorer (**24/78 = 31%**, which is what silently runs on no key / disabled / daily cap / timeout / HTTP or parse failure / batch overflow). **Do not train yet** — 800-odd labels gives a holdout too small to grade (SE ±0.032); the pre-registered plan is ~4,000 deduplicated headlines over ≥3 weeks, time-split with near-duplicates clustered first (the same story is re-titled by several outlets, so a random split scores memorisation), headline text ONLY (the teacher never sees `description` — `_llm_headline_scores` sends titles), graded against the golden set recoverable with `git show 1cebebf^:backtest/research/llm_news_golden.csv`. When it ships it slots in as DeepSeek → model → keyword and stays display-only. Training happens offline in `venv/`; a Module 32 test now AST-walks the root, `providers/` and `pair_signals/` asserting no `sklearn`/`scipy`/`lightgbm`/`torch` import and that none of them reach `requirements.txt` — that guard did not exist before.
+- 2026-09-09 replay recording shipped (`backtest/replay_record.py`, `backtest/replay_scorecard.py`, Module 33, 498 tests; nothing in the signal core changed, books byte-identical — `backtest/research/replay_recording_2026-09-09.md`). `--record DIR` stores every ACCEPTED trade's decision evidence in Parquet: each indicator's vote / value / confidence / BASE weight / regime multiplier / EFFECTIVE weight / contribution on the primary AND both confirmation TFs (`indicators.parquet`, long; absence = the indicator did not fire, never a zero row), per-TF regime/composite, penalty tags, setup, outcome, and the Forex Factory calendar events from 4 h before entry to exit (`events.parquet`, all-day rows matched on the New-York date; scalar `ev_*`/`us_holiday` columns on `trades.parquet`); `--record-bars` adds one float32 row per ANALYZED bar with the same votes plus 1 h/4 h/24 h forward moves in ATRs (`bars.parquet`). The manifest carries the engine + gate fingerprints, the regime weight table in force and the book MD5. Calendar source is the gitignored `backtest/cache/calendar.parquet` (15,124 FF events 2023-09-01→2026-09-10, unix-UTC `dateline`; `timeLabel` is display-only); the scraper that built it is back at `scripts/ff_calendar_scraper.py` (resume + idempotent merge, ~1.6 s/request). The 2-year books (2024-09-05→2026-09-05, $0.65) live in `backtest/research/replay_2026-09/{m5,m15}/`: M5 781 / +146.07R / PF 1.52 (halves +28.0/+118.1), M15 793 / +117.62R / PF 1.41 (halves +38.2/+79.4); control-vs-recorded MD5 equal on both. First-look facts are descriptive only (Bollinger opposes 94% of M5 entries; divergence-opposed M5 trades did best again; every indicator's SELL votes were followed by moves against them on both TFs; the bar-level baseline drifts +1.9 ATR/24 h on M5 so BUY votes must be drift-adjusted first) — every candidate change stays a rule-1 rerun with a control book.