Live signals.db review (2026-09-04 → 09-25, 100 unique setups replayed on the
scans' stored candles): the stop-hunt claim is refuted (71% of stop-outs ran
≥ 1R past the stop, the same as the books' 70–72%); −25.1R of the −37.8R raw
live book were consecutive-bar re-entries the backtester's 5-bar cooldown never
takes; the cooldown view sits at the 3rd–5th percentile of the books' three-week
windows.
Delivery (book-neutral):
- telegram_bot: 5-bar alert cooldown per timeframe (_ALERT_COOLDOWN_BARS =
backtest.cooldown_bars); setup block prints the trail first as the tested
exit, TP1/TP2 as guides with their R, "Planned TP1 RR", and one validated-book
line per setup class (agent_panel_base_rates.json); the informational 100-pip
TP line is removed (owner request).
- AI block rewritten for a new trader: "AI check: the odds look better than /
worse than / about usual", the chance "in 100" against the usual odds, one
reason for and one against, no verdict word or confidence figure; rendered
whenever the verdict carries a probability. agent_panel: persona prompts ban
trading jargon (PANEL_PROMPT_VERSION v2.1-2026-09-26); the macro-source
verdict now carries its persona's p_win (every live v2 row had
panel_probability NULL). Dashboard and guides mirror the wording.
- telegram_bot: level-breach rescans skip the settlement-pause window (the
20:45 M15 bar was re-logged at 22:01/22:06/22:11 every evening).
Signal (pre-registered rule-1 campaign, CUDA twin paired with gpu_backtest.py):
- M15 grade-A SELLs blocked (sell_min_quality ["A+","B"], M15 override 6):
W2 813/+154.15R PF 1.53 vs 931/+146.14R PF 1.43, OOS +68.5 vs +60.67,
W3 max DD 20.9 vs 19.3R. PASSED a–f.
- Printed target 1 floored at 1R in every generator (signals.tp_print_floor_rr),
applied AFTER the min-RR and spread gates so no trade changes; mirrored in
cuda_signal_kernels.py K2. Label-only re-pin of all four timeframes (only
tp1/rr_planned columns move). A floor applied before the gate cost −24.6R and
was rejected.
- FAILED: M15 min_rr_ratio 1.0 (trade-identical, −2.0R). Owner's call: M5
tp_floor_rr 1.0 (+48.9R W2, max DD 10.1 → 15.9R), not shipped.
New pins: M5 W2 e599c8a8…, M15 W2 5b65ecf6…, M15 OOS 09b8a8e7…, H1 W3
a44fd19b…, H4 W3 f3587bae…; twin and reference MD5s equal on every window.
Base-rate table rebuilt from the new pins. Suite 558 passed.
Evidence: backtest/research/rr_and_sell_campaign_2026-09-26.md (plus the
2026-09-18 loss-anatomy and wrong-direction reports it builds on).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Agent panel v2: personas return p_win anchored to the book base rates with a
code-computed decision context; verdict = equal-weight mean (verdict_source
"mean"); only the macro persona runs live (agent_panel.personas ["macro"]);
prompt version and exact payloads logged; offline replay harness
scripts/panel_replay.py; base-rate table agent_panel_base_rates.json
(scripts/build_panel_base_rates.py). Backtester `strategy` column now carries
the generator's own stamp (label-only; five reference books re-pinned, all
62 other columns identical).
News corpus (after evaluating the owner's ML-NEWS scikit model against
DeepSeek: golden set 39/78 vs 78/78, out-of-sample agreement 64% against a
90% self-agreement ceiling; verdict: fallback only, not a replacement):
- NewsSentiment.xau_score + archive column xau_score keep DeepSeek's gold
score for every headline (930 of the first 1,546 DeepSeek rows had none);
an existing newsdata.db gains the column in place on first open
- headlines naming no tracked currency are scored and archived
(currencies_json []) but never returned to the platform
- _apply_llm_sentiment sends every new headline in its cycle, 25 per call;
a failed call ends the cycle's sending
Dashboard publisher never writes a database: main() calls
app_config.set_db_writes_for_process(False), honoured by signal_log_enabled()
and news_archive_enabled() (owner decision: the bot is the single writer).
Test suite: the hermetic app config now points both writers at a
suite-private temp dir — the RSS fixture tests were writing fixture
headlines into the real logs/newsdata.db — with a Module 32 test pinning it.
529 tests.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Signal path
- A shipped setup always names a direction: grade C needs confidence
>= 45 (gate 6c grants B at >= 45) while BUY/SELL needs >= 50, so C/B
reports reached the alert gate as NEUTRAL with a full setup. The
backtester traded them (the replay never reads final_signal) while
Telegram rejected NEUTRAL first — 5/91 trades, +3.7R of +22.2R, in
the H1 12-month book. _attach_core_setup stamps the setup's direction
and repairs the pre-gate action text; no-setup reports stay NEUTRAL.
- analyze_pair fails the scan when ANY required frame is missing
("DATA UNAVAILABLE:", scan_kind='failed'); a lost confirmation frame
used to run with an unearned -15 MTF penalty.
Outcomes and labels
- advisory_outcome.resolve_locked_floor: a gap the database cannot fill
resolves at the locked trailing stop (GAP_LOCKED, result.floor) when
the stop had crossed the entry; the rebuild fires on both gap reasons
(rows 915/920: +1.14R / +0.84R floors; resolved sum -2.46R -> -0.48R).
- label_scan_outcomes: SCAN_LABEL_MAX_AGE_H = 72 h, stale rows never
anchor the shared bar stream, labeled_at bumped on re-examination,
_stream_since warns at its row cap (the Labor-Day pile would have
blinded scan_outcomes around 2026-09-14).
Providers and scheduling
- The daily settlement pause (Mon-Thu 17:00 New York) is a scheduled
closure for _has_latest_closed_bar (whole-span label walk-back) and
_is_data_fresh (closed seconds are not age); bot and dashboard skip
scans whose just-closed label sits in the pause (61 false stale rows
in three nights).
- fetch_range retries a failed page in place (three failures on one
window raise) and logs an ERROR for interior gaps of >3 unscheduled
bars.
- pd.Timestamp.utcnow() replaced in providers/base.py and engine.py;
requirements pin pandas<4, numpy<3.
- telegram_bot: boundary loops share one scan per bar with the breach
handler (_remember_scan keys on the analysed bar; breach defers 20 s
after an M15 boundary); the rotating file handler attaches in main()
only (the test suite used to write into logs/telegram.log).
- dashboard_publisher: catch_up_timeframes re-scans boundaries missed
while a cycle overran or the process was down.
Backtester and engine guards
- _replay_span returns the per-span GPU cache-miss count; run_backtest
sums, logs it (must be 0) and writes cache_fallbacks to the manifest.
- regime_weight_table raises on a partial regime_weights_override.
Verification: 519 tests; M5/M15 2-yr, M15 OOS, H1/H4 3-yr books
byte-identical before and after; live signals.db untouched (checked on
a copy). CLAUDE.md carries the dated entry and the blocked-hours union
semantics that every book was validated under.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0139xz1J6YccndA2be6yFnEo
Delivery-side fixes from the 2026-09-08 three-day signal-log + telegram-log
review. Nothing in the signal core changes: the 3-year H1 and M15 books at
$0.65 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.
- signal_aggregator: step 7 moves into `_attach_core_setup`; a final pass
after the PB/CP archetypes (`_drop_stale_spread_gate_vetoes`) moves a core
generator's "SPREAD GATE: ... costs would eat the edge" veto to
`cleaned_warnings` whenever a later generator shipped a setup. The
2026-09-08 07:45 UTC A-grade alert carried that line beside a $26.49 TP1;
a fifth of H1 and a third of M15 A-grade alerts had been doing the same.
The prefix is now the `_SPREAD_GATE_PREFIX` constant shared by both
emitters and both cleanups.
- telegram_bot: alerts require a trade setup (`_is_alert_candidate`, shared
by the scan and breach paths; no-setup candidates log
`alert skipped: ... (gate reason)`). The 07:50 UTC B-grade SELL with no
setup had gone out under the same warning.
- telegram_bot: every delivered alert / breach alert / violent-move heads-up
logs an `... sent:` INFO line with the signal-log row id (only failures
were logged before, so the review had to reconstruct sends from panel
rows). Send fan-out is one `_broadcast` helper for both paths.
- telegram_bot: `telegram.log` rotates daily x 30 instead of 3-hourly x 240
(old `_HH` files on the live box need one manual delete).
- signal_log / CLAUDE.md: `cleaned_warnings_json` is no longer a CP-only
marker; docs say so.
- test_all: 468 tests (+2): stale-veto cleanup incl. the PB-shaped case and
the both-vetoed case, and the alert-candidate rule with its skip line; the
rotation test asserts MIDNIGHT / interval 86400 / 30 backups.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018xY4NPPrx6NJoG6QWHzwpd
2026-09-07 signal-log inspection follow-up — three fixes that do not depend on sample size.
- providers/base.py: the signal-side 10σ outlier filter is now a bad-print filter. A >10σ
close-to-close bar is dropped only when the NEXT close snaps back ≥80% of the way
(OUTLIER_SIGMA / OUTLIER_REVERT_FRACTION); persisting moves and the newest bar are kept.
The unconditional rule deleted ~36 real M5 news bars/yr in the backtester (whole-frame σ)
and a different ~24/yr live (fetch-window σ) — e.g. the 2026-09-04 12:30 UTC NFP bar
vanished from M5+M15 and the 12:35 scan was a byte-identical repeat of 12:30. 3-yr Oanda
drops are now M5 2 / M15 1 / H1 0.
- providers/oanda.py: transient 401/5xx/network faults retry along RETRY_DELAYS_S=(2,5,10)s
under RETRY_BUDGET_S=45s. 11 boundary scans in two days had lost their primary frame at
xx:00:22–:26 UTC; one 2 s retry could not outlast the top-of-hour blip.
- signal_aggregator/_log_signal: a scan whose primary fetch failed logs scan_kind='failed'
(price 0, no signal); analyze_signal_log.parity_report skips it like 'blocked'.
- test_all.py: 466 tests (+6): retry ladder walks all delays, 5xx blip recovers on the third
attempt, retry budget stops the ladder, persisting/newest/partially-retraced 10σ bars kept,
failed-scan row kind, parity skip.
- CLAUDE.md: cleaning rule, retry ladder, failed scan kind, dated current-state note.
Book impact at $0.65 (rule 1 still passes everywhere, both halves positive):
12mo M5 557/+112.21/PF1.56 → 560/+108.81/1.54 M15 397/+73.93/1.53 → 398/+74.61/1.53 H1 identical
3yr M5 857/+161.3/1.53 → 870/+153.6/1.50 M15 1086/+148.7/1.37 → 1092/+151.9/1.38
H1 255/+70.5/1.98 → 253/+70.0/1.98 H4 113/+22.7/1.63 → 114/+21.7/1.59
Only trades on the previously-deleted event bars move; the pinned M5/M15 CSVs are no longer
byte-reproducible and should be regenerated at the next re-pin.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDnBxJ2y8GRUM6ZNFJs1hZ
- SignalAggregator.pb_funnel / cp_funnel and the 19 _reject() ticks inside
_try_m5_pullback / _try_chart_pattern: class-level rejection counters nothing
in the repo reads. Every gate keeps its early return; only the counter bump
before it is gone.
- _TP_QUALITY_SCALE: the "B"/"C" entries and the 0.75 default could never be
selected — the XAU ATR generator is only called for A+/A. The lookup is now
a direct index so a violated invariant fails loudly instead of scaling silently.
- engine._classify_regime no longer takes the unused DataFrame argument, and the
structure S/R branch tests the 5-minute frame only (M1 is never a primary).
Parity: 12-month M5 and H1 books regenerated byte-identical
(md5 a28fed4becd056466c26d6a947e1a0e5 / d20405ef64f069770c258bb9bed1ac90).
Suite: 461 passed, 0 failed.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0156YhogLytzj9P2iH1skGbL
- analyze_pair(context=...): no caller ever supplied it (main, telegram_bot,
pair_signals, tests), so the value was always None; the parameter and the
pass-through to analyze_from_dataframes go (analyze_from_dataframes keeps
its own context parameter).
- The three specific "SELL FILTER: quality|volatility|regime" stale-prefix
entries were shadowed by the general "SELL FILTER: " entry in the same
startswith() tuple; every emitter carries the general prefix.
- chart_patterns: the detectors never read the Open series, so the `o`
parameter is dropped from the three detectors, the PATTERN_DETECTORS
lambdas and the dispatcher.
- Corrected the pb_funnel/cp_funnel comment: nothing reads those counters
(the "read by backtest tooling" claim was stale); they stay as cheap
diagnostics.
Parity: 12-month M5 and H1 books regenerated byte-identical
(md5 a28fed4becd056466c26d6a947e1a0e5 / d20405ef64f069770c258bb9bed1ac90).
Suite: 461 passed, 0 failed.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0156YhogLytzj9P2iH1skGbL
New additive columns on every logged scan: bar_ts (primary closed-bar label, indexed with
timeframe), engine_fingerprint (hash of the scoring config incl. the regime weight table in
force - SignalAggregator.engine_config_fingerprint, twin of the gate fingerprint), code_version
(short git SHA), penalty_tags_json / penalty_total / momentum_fading / exhaustion_count, and
post-hoc watch_level / watch_kinds / watch_line written by violent_watch.attach_watch via
signal_log.update_watch. Each indicators_json entry now also carries eff_weight (base weight x
regime multiplier, the weight the composite actually used) and wscore.
engine: IndicatorSignal.effective_weight / weighted_score are stamped inside _compute_composite
(record-keeping only); regime_weight_table(config) is the single source of truth for the
multiplier table (override honoured). Composite output byte-identical (120-report hash proof).
New table scan_outcomes: model-free forward moves per scan (fwd_1h/4h/24h in $ and ATRs,
mfe/mae over 24h, burst_1h, bars_used, complete) measured on the stored candle stream with the
advisory model's continuity rule; labelled opportunistically at the end of log() (time-boxed,
throttled, idempotent) - no daemon; scripts/label_scan_outcomes.py drives the backlog
(--dry-run is read-only). analyze_signal_log --scan-outcomes: forward move + hit rate by
TF x signal/grade with a NEUTRAL baseline and burst rate by watch level. Readers dedup by the
first row per (timeframe, bar_ts).
Module 31 (10 tests). Suite: 463 passed. Backtester untouched (test-locked).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RWV2yLt3YXHHGVYHSQU2EE
New violent_watch.py assesses each scan from the cached Forex Factory week events, the
UTC clock and the report's own indicator states, and stamps SignalReport.watch_line
post-hoc (telegram alerts, CLI, dashboard). A standalone heads-up goes to the owner on
the fastest scanned timeframe when the level is "hot", once per UTC hour per trigger
set, capped per day. Config block `violent_watch` (enabled False, alert_line, headsup,
max_headsup_per_day, headsup_min_level); env VIOLENT_WATCH_ENABLED.
Rule chosen from a 3-year bar-by-bar sweep (backtest/research/watch_rule_eval.py,
recorded in sizing_campaign_2026-09.md): hot = high-impact USD release within 30 min
(before or after) OR a hot clock slot (weekday 12-14 UTC, 01 UTC, Sunday 22-23 UTC)
with at least one price-extreme reading (outside Bollinger bands, RSI <30/>=70, outside
Keltner); ~8.6% of bars, next-bar odds 3-5x normal, ~4 hot hours/day. watch = clock
slot alone or two price-extreme readings (~2x). Context readings (ADX rising, extreme
volatility, transitioning, MACD acceleration) are phrases only - letting them raise the
level flagged 27% of bars at ~2x and would have sent ~13 heads-ups/day (rejected).
Direction sentence: DOWN when bearish readings dominate (70-78% on both TFs), UP on M5
only. Never a gate or exit: every open-trade protection tested lost money.
Test-locked out of analyze_from_dataframes / engine / backtester. Module 30 (18 tests).
Suite: 453 passed. Guide PDFs not rebuilt.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RWV2yLt3YXHHGVYHSQU2EE
Walk-forward "second opinion" stake multiplier (meta-labelling on 65 decision-time
features, 3yr Oanda books @ $0.65) FAILED the pre-registered rule on M5 (clean null)
and M15 (lift inside the label-shuffle null, explained by BUY-vs-SELL). Nothing that
changes numbers ships; the 12-month M5 book re-run is byte-identical to the pinned one.
Behaviour-neutral plumbing:
- position_sizing.build_features (65 features; multiplier path returns 1.0, no artifact)
- SignalReport.penalty_tags / penalty_total / momentum_fading (record-keeping)
- TradeSetup.stake_multiplier = None (reserved)
- gpu_backtest --export-features (default export stays the pinned 63 columns; extras
forced to the tail even when the first trade lacks an indicator)
Research tools (backtest/research): book_report, burst_map, burst_precursors,
open_trade_bursts, calendar_impact, burst_indicator_map, sizing_model, watch_rule_eval.
Findings recorded in backtest/research/sizing_campaign_2026-09.md: M5 stretch measures
are positive in both halves; every open-trade protection against violent bars loses on
both TFs; high-impact USD releases cause 20-31% of violent bars but trades held through
them do better; violent bars come when price is already at an extreme, not from calm.
Module 29 tests (13). Suite: 435 passed.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RWV2yLt3YXHHGVYHSQU2EE
agent_panel.verdict_source defaults to "macro": the macro take's stance is the
verdict, the synthesizer call is skipped (4 LLM calls per panel instead of 5);
"chair" restores the four-voice consensus + OPPOSE guard. Verdict source is
cached per mode and logged (panel_verdict_source). 406 tests.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FC4s6VCURFte4dub4xKxYL
New signal_log.py logs each SignalReport to logs/signals.db via a
fail-safe hook in SignalAggregator.analyze_pair — the live-only entry
point used by the bot, daemon, CLI and dashboard. Backtester calls
analyze_from_dataframes directly, so live==backtest parity is preserved
(test-locked). Rows carry signal/grade/price/spread, MTF + trade-setup +
reasons/warnings JSON, and news-at-scan-time (calendar events + scored
headlines JSON via NewsEngine class-level caches).
Config: signal_log block (enabled, path; env SIGNAL_LOG_ENABLED/PATH),
source tagged per consumer (telegram/oanda_daemon/cli/dashboard).
381 tests pass (24 modules).
ai-hedge-fund-style panel: 4 DeepSeek personas (macro, technical, risk,
contrarian) review a finished SignalReport concurrently on a
panel-private thread pool, then a synthesizer chair issues
AGREE/NEUTRAL/OPPOSE + confidence + rationale. Advisory/display-only by
construction: consumers attach post-hoc via attach_panel; the fields
live on SignalReport with empty defaults and analyze_from_dataframes
never touches them (backtest parity, locked by test).
Spend control on top of the shared 500/day client cap: max_panels_per_day
UTC budget (default 20), exact per-bar result cache, and a reuse TTL
keyed (symbol, tf, direction) so a persisting signal re-serves its
verdict across bars instead of burning 5 requests per bar. <2 valid
takes skips the synthesizer; any failure degrades to no panel.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SisdfkxnXNL9iZsXxN3T15
Owner decisions from the 2026-08-04 audit report:
- spread-in-exit-triggers: DECLINED (books stay comparable; ledger in
CLAUDE.md records the measured ~4-6% optimism).
- weekend clock: FIXED. The market week is Fri 17:00 -> Sun 18:00 New
York time, not Fri/Sun 22:00 UTC — during US DST the Oanda feed stops
at 21:00 UTC (measured: 159/162 summer weekends). All three weekend
implementations (providers/base, signal_aggregator, gpu_backtest
calendar filter) now share the NY-anchored rule; a new test sweeps
them hourly across both DST transitions and asserts agreement. No NTP
needed — the host clock was fine, the constant was wrong.
Ends the weekly false-"STALE DATA / feed lagging" hours (summer Fri
21-22 UTC, winter Sun 22-23 UTC).
- FB2_EMA_LO (delegated): kept as a research knob but crash-proofed —
an unparseable value degrades to the tuned default 0.5 with a warning
instead of killing every scan. Documented in CLAUDE.md.
Book gates: M15 12mo pinned book byte-identical (+78.7111R, 397
trades). M5 12mo drops exactly ONE trade (556 -> 555, +97.56 ->
+98.64R): a phantom whose signal fired on the final summer-Friday bar
(decision 21:00 UTC = NY 17:00 market close) and was "filled" at
Sunday's reopen 49 hours later, -1.08R — an entry live trading can
never take. The backtest no longer contains it; live == backtest
parity improved. M15 unaffected because hour 21 is in its blocked set.
Suite: 341/341 green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UnZkzeFcTu9agFg8BDh2tn
Three display/warning-only upgrades to the news engine, all behind the
existing keyword path as fallback — signals are never gated on LLM
output (event_risk_caps_signals / sentiment_affects_signals stay False
per the 2026-07 audit; backtester use_news=False untouched):
- Headline sentiment: one batched call per 600s cache window (capped at
25 new headlines, sha1-cached so repeats are never re-sent) overrides
the keyword scores per currency with the gold-inverse doctrine encoded
in the prompt.
- Release surprises: get_event_risk_level's description now says which
way the print pulls GOLD ("— gold bearish: strong jobs lift yields"),
fixing the documented numeric-only gap; risk level and windows
unchanged.
- Market summary: 1-2 sentence macro blurb on SignalReport.news_summary,
rendered in Telegram alerts (📰) and the CLI (🧠).
New llm_client.py: requests-only OpenAI-compatible client (no SDK), one
attempt per call, UTC-day request cap, singleton so the cap survives
per-scan NewsEngine reconstruction. Config: deepseek block in
config.json + DEEPSEEK_API_KEY env; absent key = fully inert.
Live-measured API behavior baked in: V4-Flash is a hybrid thinking model
— requests send thinking:disabled (reasoning otherwise blew the timeout
and could exhaust max_tokens before the answer); prompts are static
module constants so DeepSeek's prefix cache prices repeat calls at the
cache-hit rate; scores use compact [s, xau] pairs; timeout 12s.
Tests 273 → 286 (new Module 19 + deepseek config tests).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nsd8HcrV67GbCZXxZEn1Ma
Gate 6h (big-candle entry gate): blocks all XAU entries (core, XAU_PB,
XAU_CP) when the primary TF's last 24h contains a bar with range >= K x
prior-bar ATR. Ships default-OFF: validation rejected every threshold
(K 3-6 on M5/M15 12mo) — removed trades were net-POSITIVE at every cell;
M5 quiet-window trades are the only losing bucket. Config knobs
big_candle_atr_multiple/lookback_hours, --big-candle backtest flag,
Module 17 tests.
M5 relaxation SHIPPED (validated 2026-07-30, Oanda bars, $0.30 spread):
m5_min_abs_score 30->25, m5_min_confirmation_score 10->5, hour-21 UTC
unblocked M5-only. Beats control in ALL 6 calendar years of the 5yr
replay (+230.9R/2319 PF 1.26 DD 21.3 vs +168.0R/2048 PF 1.21 DD 27.0),
12mo +107.4R PF 1.36 vs +84.6R PF 1.33, OOS +26.1R vs +17.2R, $0.70
spread +86.9R vs +66.4R. M15: every relaxation retested negative —
config unchanged. Declined-but-open: M5 hour-22 block, M15 hour-16
unblock (both fail 5yr/OOS). Evidence in the factory docstring.
Tooling: generic --set PATH=JSON research override in gpu_backtest
(typo-proof, logged); diagnose_filters filter_tags refreshed for gates
6d-6i with a staleness-guard test (Module 18); research scripts
hour_analysis/burst_profile/book_diff. Suite: 268 tests.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014jqHpmNYmwYMowF6bPZyeD
One class per provider (TwelveData, Oanda, Tiingo), all subclassing
BaseDataProvider which owns cache, validation/cleaning, freshness +
boundary guarantees, fetch_range pagination, and streaming stubs —
subclasses implement only _fetch_raw. config.json data_providers.active
selects the live feed; data_provider.py / oanda_provider.py stay as
back-compat shims. Module 16 provider-abstraction tests (253 total green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Mc7XKYyPySVXiUpcN82QEK
Five-agent audit of the signal path; all fixes are delivery/measurement
side. Regression-proven behavior-neutral on validated signals: the 5yr
M15 replay reproduces the pinned book exactly (+210.65R PF 1.35
WR 57.8% MaxDD 33.90R). Tests 215 -> 221.
- TradeSetup carries trail_arm_r/trail_dist_r stamped from the producing
config; Telegram trail plan reads them (was showing default 0.6/0.75
instead of the validated M15 arm 0.70 / M5 dist 0.85)
- Alert dedup: fingerprint = (signal, quality, direction); entry movement
compared against last-ALERTED entry/risk (old SL-dist bucket churned on
ATR drift -> near-duplicate alerts every candle)
- Fresh fetches enforce the closed-bar boundary check with one 3s grace
retry, then warn (feed lag no longer silently analyzes the prior bar)
- Backtest max DD includes the zero baseline (opening losing streaks
were undercounted; pinned M15 DD unchanged)
- Replay loads the regime-filter TF when the H4-ADX gate is on (was a
silent no-op for M5/M15 primaries)
- clear_cache under _cache_lock; dead ROC zero-cross branches removed
from both engines (provably unreachable); RSS dates via
parsedate_to_datetime (named zones no longer stamped UTC); event risk
grades from the soonest event; .env inline comments stripped;
auto-scan skips NEUTRAL like the breach path; report_to_dict honors
the weight-0 skip rule and emits strategy/trail params; PB funnel
counts spread_gate/accepted; session-filter fail-open paths log loudly
- Docs corrected to match validated code (H&S neckline = lower peak,
exhaustion anchors farthest tracked swing) with re-validation warnings
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R5soRUeRBi8kHMB99csrJy
Resolves the Twelve Data API key through app_config.provider_api_key()
(env → config.json → no default) instead of a hardcoded field on
PlatformConfig. Adds data_provider.PROVIDER_REGISTRY + make_data_provider()
so signal_aggregator and the backtester build the active provider through
one seam; telegram_bot.py gets the same one-line WS-key fix now (per Task 4's
already-planned edit for that line) so the suite doesn't crash on import
between this commit and Task 4's full wiring pass.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VPNF37MP9Kuv62B1VBrZ1b
Add NewsConfig.sentiment_affects_signals (default False), gating the +5
aligned-sentiment confidence bonus and the 10-pt news_conflict penalty.
With event_risk_caps_signals already False, live decisions are now pure
technical, matching the backtester (use_news=False). Warnings and the
sentiment display are unchanged. Adds a gating regression test.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FMaaPdrrrQuh1VvGQH1icN
Third entry archetype: completion of classical chart patterns (double
bottom/top, H&S top / inverse H&S, bull/bear flag) — the tight canon
with the strongest documented stats. New chart_patterns.py detector
module (pure functions, ATR-unit geometry, closed-bar slices only,
first-close-beyond trigger discipline so a breakout fires exactly
once and never repaints).
Pipeline step 10 in analyze_from_dataframes: fills only empty setup
slots, allows weak-composite F grades (normal at reversal completion)
but inherits session/day blocks, mirrors MTF-opposition + SELL
vol/hour + M5 protections. Geometry: SL = pattern anchor -/+ 0.5 ATR
(min $8), TP1 = measured move, $9 spread gate, trailing exit does the
real work. Strategy tag XAU_CP.
cp_enabled defaults to False everywhere — per-TF adoption happens in
pair_signals/xauusd.py only after the standard validation rule passes.
Backtester: --cp / --cp-patterns flags to test the candidate, XAU_CP
tag recognition.
Suite: 183/183 (21 new tests: detectors, trigger discipline,
invalidation, gates, pipeline, default-off safety).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015kby7M9JwpYi8NQT2yoxAi
Fixes all high+medium findings from the 2026-07-16 full-codebase audit.
High:
- data_provider: cache now rejects snapshots missing the latest closed
bar (_has_latest_closed_bar, weekend-exempt). The wall-clock TTL served
mid-candle snapshots at the next boundary, so live M15/H1 scans ran one
full bar behind and the bar-close session check evaluated the wrong hour.
- signal_aggregator: gate 6e (SELL quality) only acts on actionable
grades. It was re-grading session-blocked F reports to D, which
re-opened the M5 pullback path (entry check: quality != "F") on
session-blocked bars.
- gpu_backtest: port the live EMA-ribbon compression/expansion logic
(+10 conf on expansion) into the GPU engine. Its absence diverged
composite score/conf on 7-9% of bars and flipped signals on ~0.2%.
Medium:
- gpu_backtest: Bollinger precompute uses the exact pandas rolling calls
(cumsum E[x2]-E[x]2 lost precision at gold price levels; bb_position
error up to 7.6e-3 crossed signal thresholds, worst during squeezes).
- gpu_backtest: --confirm-tfs defaults to the pair factory's live
confirmation map instead of a hardcoded 1h,4h.
- gpu_backtest: entries re-anchor to the first RAW bar after the signal
bar, so >10-sigma spike bars dropped by cleaning are no longer skipped
at entry (favorable bias).
- gpu_backtest: per-TF warmup pad — D1 confirms get >=210 bars at window
start (62-day pad gave ~61 bars: no EMA200, unlike live).
- data_provider: thread-safety locks for cache + throttle (shared
provider across the bot's 4-thread pool); outlier filter keeps the
snap-back bar after a spike instead of punching a two-bar hole.
- telegram_bot: alert fingerprint gains a risk-relative entry bucket
(materially-moved setups re-alert) and is committed only after a
successful send (failed sends no longer consume the alert for 1h).
- engine: disabled_indicators=["macd"] now removes the MACD pillar from
scoring (histogram still computed for divergence).
Verification: test_all.py 187/187 (new Module 11: 8 audit-fix
regressions incl. a live-vs-GPU parity sweep with ribbon coverage);
real-data parity sweep over 813 slices of 14mo XAU M15: zero divergence
(pre-fix: 7-9% of bars diverged); both original bug repros no longer
reproduce; cached end-to-end backtest run passes with the live confirm
map. Books shift slightly vs pre-fix runs — comparability note added to
CLAUDE.md; pinned tuned numbers need re-validation under the fixed
backtester.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Fix 1 — mixed-"trending" regime (ADX strong, EMAs unaligned) was
hard-coded bullish in 5 sites: RSI reinterpretation used the uptrend
playbook (RSI 38-55 -> STRONG_BUY conf 75), BUY-only confidence
bonuses, BUY-only extension penalties, no counter-trend penalty.
Now: extremes-only RSI (as transitioning), no directional bonuses,
symmetric extension penalties. 6yr validation: H1 +67.2->+61.9R
(mixed skew 16B/11S -> 4B/7S), H4 +31.6->+27.5R (the 4 bias BUYs),
M15 +152.6->+154.7R, M5 +207.1->+206.1R with maxDD 28.7->24.3R.
The ~8R aggregate give-back is bull-market subsidy from the bug.
Fix 2 — live session filter now applies the EXACT replay-loop
predicate (session window + blocked hours on the signal bar's CLOSE
time, from the bar timestamp not wall clock). Replaces a wall-clock
filter gated on a "XAU STRATEGY" marker that no code writes, which
made live diverge from every validated book.
Fix 3 — delete dead _check_xau_oversold_bounce and
_generate_xau_trade_setup (unreachable; the latter hardcoded BUY and
skipped the spread/RR gates).
Backtester: parallel bar replay via forked workers (--workers,
default auto = cores-2 capped 12). Per-bar evaluation is pure; the
entry cooldown is applied in an ordered phase-2 pass. Proven
byte-identical to sequential on the same window; ~300 -> 1030-1680
bars/s (6yr M5: ~26min -> 5-8min). --workers 1 = sequential path.
Tests: 179 passed (new mixed-regime RSI regression test).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
M15 mixed-"trending" regime block (ADX strong, EMAs unaligned =
momentum without direction): -6.8R/24 trades on 12mo, negative both
halves; blocking scores +65.7R/499 PF 1.38 DD 10.4R vs +59.0R/521
PF 1.32 DD 12.6R, better both halves, OOS 2024-06->2025-03 neutral
(-1.4R = the 4 such trades there). M5 never trades this regime.
New XauStrategyConfig.blocked_regimes / sell_blocked_regimes knobs
(default empty) + aggregator gate 6g2; M15 factory opts in.
M5 hour-9 UTC block (post-London-open whipsaw): -28.9R/143 over the
6-year book, negative 5/7 calendar years. Full-history replay with
block: +207.1R vs +178.3R, max DD 28.7R vs 35.5R, 2023 -13.0R ->
breakeven; tuned 12mo book unchanged (139.0 vs 139.3). Hour 9 is
positive on M15 (+1.7R 6yr, +8.6R OOS) so the block is factory
M5-only.
Rejected after validation (documented in xauusd.py + session memory):
transitioning-SELL block (bull-market artifact, +5.7R in 2024 OOS),
M15 trending_down block (M5 cross-TF contradiction), losing-streak
circuit breakers, 30d efficiency-ratio trend gate, and all entry-level
indicator filters (winners == losers at entry on every feature).
Full-history context (Twelve Data 5min starts 2020-03-16): M5 6.3yr
+207R PF ~1.3 post-fix with losing year 2021; M15 +152.6R PF 1.19
DD 49.7R — edge is regime-dependent, chop years are structural.
Tests: 178 passed (new gate test + factory-equivalence updates).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Replaying the live danger/caution gates against the validated 12mo XAU
M5 book showed they suppress 25-35 trades worth +4.6..+7.2R/yr — trades
entered near USD releases outperform the book average (+0.475R vs
+0.256R). New NewsConfig.event_risk_caps_signals (default False =
XAU-tuned) keeps EVENT RISK warnings in alerts but no longer guts
confidence; EUR factory re-enables the caps (evidence is gold-specific).
Pinned XAU M5 regression: 547 trades, +139.94R, trade-identical.
Tests: 177 passed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The SELL quality gate (6e), SELL volatility gate (6g), M5 noise gates
(6h), ATR trade-setup generator, and M5 pullback archetype are all
gold-tuned but previously keyed only on xau_strategy.enabled or the
timeframe, so they would have silently applied gold constants to any
other pair. Each now also requires display_name == "XAU/USD".
No behavior change for XAU (conditions are tautologically true for the
only registered pair today) — verified by test suite and pinned-window
backtest CSV diff.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Second entry pattern alongside the Keltner-breakout: enter on the
retrace to EMA21/KC-mid in an established trend (EMA >=3/4 aligned,
ADX>=20, retrace 0.8-6.0 ATR, resumption bar closed back in trend
direction, RSI in the pullback zone not the extended zone). Detector
lives in the aggregator so live and GPU-backtest paths share one
implementation. Own gates (composite floor 15, confirmation-TF
agreement, MTF opposition, vol regimes) with a diagnostic funnel.
PB stops anchor to the pullback swing (~$8 vs ~$25 breakout), so PB
trades get their own trail arm (0.8R vs 0.6R global) via per-trade
trail overrides in the backtester and strategy-aware Telegram alerts.
12mo M5 validation: PB alone +38.3R/254 trades, positive both halves
and all four full quarters; core breakout trades undisturbed
(+101.9R vs +105.0 baseline). Combined book: 549 trades (~10.6/wk),
+140.1R, PF 1.78, DD 11.3R vs 311 trades/+105.0R before.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Gates 5-6 in the M5 filter: block 'high' vol regime entries (chop
transition — 61% WR, negative total while low/normal/extreme all
positive) and block SELLs with RSI<25 (dip already extended, mean-
reverts before target). 365d: 419→311 trades, +41→+48.4R, DD 11.2→8.0R.
Replace the fixed RR-0.5 take-profit with a trailing stop: arm at
+0.6R favorable excursion, trail 0.75R behind best price, stop levels
update at bar close (conservative). The RR-0.5 target was the binding
constraint — winners kept running after TP. Whole trail parameter grid
(arm 0.4-1.2 × dist 0.5-1.25) lands +90-103R, a plateau not a spike;
also +91.7R on the pre-gate 419-trade population. TP1 stays advisory;
Telegram alerts include the arm price and trail distance. max_hold
200→300 bars (trailing exits need room; +10R).
12mo M5, same 311 trades: +105.0R total, PF 2.27, DD 7.9R, 61% WR,
train/val +54.6/+50.4, every quarter positive.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
ATR was computed after EMAs so ribbon compression detection was dead code
on all timeframes — moved ATR first. Extended M5 swing S/R lookback from
50 to 120 bars. Added four M5-specific filters: min score, MTF agreement,
confirmation TF conviction, and SELL quality cap (A-grade M5 shorts enter
at dip extremes and lose). Fixed backtester blocked-hour check to use bar
close time matching the live bot. Added trade geometry config (spread gate,
SL floor, TP floor RR) and primary_tf mismatch guard in backtester.
Backtest: 1133 trades/+11.5R → 419 trades/+41.0R, PF 1.03→1.37, DD 21→11R.
Train/val split positive in both halves (+19.8R/+21.3R).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Live pipeline:
- Drop still-forming candle before analysis (signals no longer repaint)
- Direction-aware MTF agreement + opposition gate (no trade against 2-of-3 TF consensus)
- SELL volatility gate: block shorts in low/normal vol regimes (config sell_vol_regimes)
- Divergence: 40-bar lookback, price/indicator swings must align within ±2 bars
(fired on 0.5% of trades before, 29% after), trending weight 2.0→1.0
- RSI regime_aware by default; transitioning regime only signals at extremes
- ADX ±DM tie no longer nudges bearish; TP2 clamped beyond TP1
Backtester realism:
- Charge $0.30 round-trip spread (--spread), gap-through SL/TP fills at bar open
- Resolve trades on RAW price path (cache stores raw bars, cleaning runs on load)
- Weekend/session filters use bar-close decision time to match live wall clock
Validated per-step on M15 2025-03-30→2026-05-31 under identical pricing:
+46.9R → +66.8R, win rate 69.9% → 73.0%, SELL side -2.3R → +9.6R.
Tested-and-reverted: ROC weight removal (-13R), BB fade suppression (-11.6R),
candlestick de-weight (-2R) — composite thresholds are co-tuned with weight mass.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The TP floor calculations used raw min_rr (0.4) but after 20% tightening
the effective RR dropped to ~0.32, failing the min_rr check and returning
None for all trade setups. Pre-compensate the floor with min_rr/tf so
post-tightening RR still meets the minimum.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Applied uniformly across all 3 trade setup functions. Lowered min_rr_ratio
0.5→0.4 proportionally to preserve trade acceptance rate. SL unchanged.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The old `weekday() >= 5` check blocked all of Sunday, but the forex
market reopens Sunday 22:00 UTC (Sydney open). Now uses proper time-based
check so the bot fetches data and generates signals on Sunday evening.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
A-grade trades entered exhausted moves with TP too far (0.75R), causing
57.7% WR and +0.25R over 30 days. Store exhaustion_count on SignalReport
and scale TP to B-level (1.0×) when A-grade has 2+ exhaustion tags.
Backtest: A-grade 73.1% WR / +3.25R, B/C unchanged.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signal daemon (signal_web.py): candle-boundary scanner mirroring
telegram_bot.py, publishes serialized SignalReport JSON to Redis.
Web API (web/app.py): thin FastAPI reader behind Apache, serves
/api/signals REST + /ws WebSocket from Redis pub/sub.
Frontend (web/static/index.html): 3-pair grid, color-coded signals,
indicator breakdown per timeframe, trade setup display.
Also: XAU-specific penalty scoping in signal_aggregator (softer
confidence penalties for gold, original strict values for forex),
pair-aware session filter routing XAU to its own blocked hours.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The general session filter ran after analyze_from_dataframes() and
wiped out XAU oversold bounce signals at most hours. Now skips the
session filter when the XAU strategy has already fired — the strategy
has its own backtested hour filter.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
BUY-only strategy for gold: enters when M15 RSI drops below oversold
threshold while price remains above EMA200 on all three timeframes
(M15, H1, H4). Uses ATR-based TP/SL (0.7x/2.0x) sized from M15 ATR.
Isolated to XAU/USD only — other pairs completely unaffected.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Drop USD/JPY from scan pairs (zero signals across all timeframes)
- Disable divergence indicator (zero impact in 14-config lab test)
- Parallelize evaluate.py with ProcessPoolExecutor (16 cores)
- Add --pair CLI filter to evaluate.py for single-pair backtests
- Add data freshness guard: stale cache auto-invalidated, primary TF
stale data suppresses signals to prevent trading on old data
- Add /cache command: shows fetch age and last bar per pair/TF
- Add /clearcache command: wipes in-memory cache for fresh API fetch
- Remove BB from WebSocket breach alerts (keep structural levels only)
- Show live prices only for WS-monitored pairs (EUR/USD, XAU/USD)
- Replace arrow emojis with text tags in indicator lines
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Trade setup TP now scales to SL distance ensuring min RR is met when
no viable swing target exists; skip swing targets with poor RR
- Add float tolerance to RR gate to prevent edge-case rejection
- M5 auto-scan runs for level cache only, no Telegram broadcast
- WebSocket limited to EUR/USD + XAU/USD (Grow plan constraint)
- Register bot command menu via set_my_commands on startup
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Expand from 2 to 4 pairs (EUR/USD, USD/JPY, GBP/USD, XAU/USD) with
automatic scanning at M15 and H1 candle close boundaries. Scheduler
staggers H1 by 60s to stay within 8 credits/min; daily budget is
480 of 800 allowed. Cache TTL is now candle-duration-aware so cross-TF
confirmation data is reused without extra API calls. Spread estimate
uses primary-TF data instead of fetching M5 separately.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Harwinder Sekhon