Sydnee nightly — entry_flow audit — 2026-06-01 P0 findings: 0 P1 findings: 5 Risks: 10 - Area: Entry Flow (Monday theme) - Bugs found (P0 / P1): 0 / 5 (1 new, 4 carried) - Risks noted: 10 (2 new, 8 carried) Cross-check: `git log --oneline -30` and `docs/strategy_decisions.md` reviewed. `docs/audit_2026-05-31_config_drift.md` (most recent audit) reviewed. All 4 carried P1s confirmed still present in current code. `git log` shows no commits since `e1981c3` that touch entry-flow time gates or the 1m BXt bias gate. --- Full report (dev branch): https://github.com/kanex1/sydnee.signals/blob/dev/docs/audit_2026-06-01_entry_flow.md Reply FROM [email protected] to [email protected] to request fixes, e.g.: "code_task on sydnee-signals-dev: apply fix for the P0 about RVOL threshold in bot.py" Sydnee Agent will propose + you APPROVE (or plain 'approve') + auto-push to dev. --- Full audit below (first 12 KB) --- # Nightly Audit 2026-06-01 — ENTRY FLOW ## Summary - Area: Entry Flow (Monday theme) - Bugs found (P0 / P1): 0 / 5 (1 new, 4 carried) - Risks noted: 10 (2 new, 8 carried) Cross-check: `git log --oneline -30` and `docs/strategy_decisions.md` reviewed. `docs/audit_2026-05-31_config_drift.md` (most recent audit) reviewed. All 4 carried P1s confirmed still present in current code. `git log` shows no commits since `e1981c3` that touch entry-flow time gates or the 1m BXt bias gate. --- ## Findings ### BUG [P1] (new): 1m BXt bias gate runs before time gates — open-30 / overnight gap poisons pending state **File:** `bot.py:5436–5452` (1m BXt bias gate) vs `bot.py:5494–5503` (time gates) **Evidence:** ```python # bot.py:5436 — 1m BXt bias gate (micro_scale block) micro_scale = 1.0 if trigger_type == "rsi_extreme": ... if _1m_against_e: if _pend_dir != raw_sig: ss._rsi_ext_pending_dir = raw_sig # ← sets pending self._log_activity(..., "RSI EXTREME PENDING: ... deferring 1 bar", "warn") return # ← exits BEFORE time gates # bot.py:5494–5503 — time gates (never reached on the return above) in_overnight_gap = 3 * 60 + 50 <= t_min < 4 * 60 in_open_30 = 9 * 60 + 30 <= t_min < 10 * 60 if in_overnight_gap: return # never reached if in_open_30: return # never reached ``` **Scenario (RTH open-30, affects prod without any toggle):** - `09:35 ET`: rsi_extreme signal + 1m BXt against → 1m bias gate sets `_rsi_ext_pending_dir`, returns with "deferring 1 bar" log. Time gate never runs. Pending state armed on a would-have-been-blocked bar. - `10:01 ET` (first allowed bar): `_pend_dir == raw_sig` (set at 09:35) → entry fires at `micro_scale = 0.5` immediately. - **Correct behaviour without the bug:** `_pend_dir = 0` at 10:01 → defers again → entry fires at `10:06 ET` at 50%. **Impact:** For any rsi_extreme + 1m-BXt-against signal at the open-30 start (09:30–10:00 ET), the first post-open entry fires one 5-minute bar (5 min) too early. Log message incorrectly says "RSI EXTREME PENDING: deferring 1 bar" for the blocked bar, hiding the real reason (open-30 gate). Same mechanism applies to the overnight gap (03:50–04:00 ET) but requires the off-hours toggle ON to produce a live trade. **Fix:** Move the time-gate block (`in_overnight_gap` / `in_open_30`) to run before the 1m BXt bias gate at `bot.py:5436`, or add an early return inside the bias gate: ```python # at top of the 1m BXt bias block (before setting _rsi_ext_pending_dir): if in_overnight_gap or in_open_30: pass # let the time gate below handle this; don't arm pending else: ss._rsi_ext_pending_dir = raw_sig ... return ``` --- ### BUG [P1] (carried from 2026-05-31): `TV_BXT_MAX_AGE_SEC` env var documented but never implemented **File:** `bot.py:3681` **Evidence:** `grep "TV_BXT_MAX_AGE_SEC" bot.py` → single hit in docstring only. `tv_bxt_signal:{sym}` Redis key has no TTL. A Friday-close signal persists through the weekend; on Monday's first loop tick the bot acts on it before any new webhook arrives. **Impact:** TV-BXt model (SNDK, real money) can enter based on a signal up to 65+ hours stale, potentially in the wrong direction vs. Monday-gap open. **Fix:** Add `EX` TTL on the webhook Redis write at `bot.py:11832–11843`, or add an age-check in `_tv_bxt_signal()` using a companion timestamp key. --- ### BUG [P1] (carried from 2026-05-31): `REVERSE_TIME_DECAY` uses `== "1"` — ignores `true`/`yes` **File:** `bot.py:5749` **Evidence:** ```python if os.environ.get("REVERSE_TIME_DECAY", "0") == "1" and _in_rth: ``` Every other boolean env flag uses `.lower() in ("1", "true", "yes")`. `REVERSE_TIME_DECAY: "true"` in YAML compose silently disables the morning size ramp. **Fix:** `os.environ.get("REVERSE_TIME_DECAY", "").lower() in ("1", "true", "yes")` --- ### BUG [P1] (carried from 2026-05-31): `daily_loss_limit` equals `capital` — DLL disabled on real money **File:** `config.json:3,6`; `core/risk.py:105` **Evidence:** ```json { "capital": 50000, "daily_loss_limit": 50000 } ``` Gate fires only after a full $50,000 intraday loss (100% drawdown). Bot live on real money since 2026-05-01. **Fix:** `"daily_loss_limit": 1500` (3% of $50K, matching `core/risk.py:RiskConfig` dataclass default). --- ### BUG [P1] (carried from 2026-05-31): `TV_BXT_TRAIL_AGAINST` default mismatch — live `1.0` vs history endpoint `1.5` **File:** `bot.py:3750` vs `bot.py:14062` **Evidence:** ```python # live path (bot.py:3750): agnst = float(os.environ.get("TV_BXT_TRAIL_AGAINST", "1.0")) # history replay (bot.py:14062): agnst = float(os.environ.get("TV_BXT_TRAIL_AGAINST", "1.5")) ``` Dashboard stop-analysis shows a 50% wider trail than what actually ran when env var is unset. **Fix:** Change `bot.py:14062` default from `"1.5"` to `"1.0"`. One-line fix. --- ### RISK (new): `in_rth` time gate (bot.py:5497) lacks weekday check **File:** `bot.py:5494–5497` vs `bot.py:3650`, `bot.py:6724` **Evidence:** ```python # bot.py:5497 — entry-gate in_rth (NO weekday guard): in_rth = 9 * 60 + 30 <= t_min < 16 * 60 # bot.py:3650 — BXt1m session window (HAS weekday guard): _in_rth = (9 * 60 + 30 <= tmin < 16 * 60) and now_et.weekday() < 5 # bot.py:5294 — RSI extreme threshold selection (HAS weekday guard): _in_rth_session = 570 <= t_min < 960 and now_et.weekday() < 5 ``` If a weekend bar somehow arrived (e.g., Polygon pushes a stale tick), `in_rth = True` for 9:30–16:00 ET, so the off-hours gate at `bot.py:5524` (`if not in_rth and trigger_type != "rsi_extreme"`) would NOT block BXt/divergence entries. IBKR paper does not send bars on weekends, making this theoretical, but the inconsistency is a latent risk. **Fix:** Add weekday guard: `in_rth = 9 * 60 + 30 <= t_min < 16 * 60 and now_et.weekday() < 5` --- ### RISK (new): `_rsi_ext_pending_dir` not persisted — restart loses 1m BXt pending state **File:** `bot.py:5447–5457`; `core/persistence.py` (absent) **Evidence:** `_rsi_ext_pending_dir` is an in-memory attribute on `SymbolState` (`ss._rsi_ext_pending_dir`). No DB column or Redis key stores it. On any bot restart (container recreate, daily rotation), the state is lost. **Impact:** A rsi_extreme that was deferred at bar N ("1m BXt against") and should fire at 50% on bar N+1 will instead re-defer on bar N+1 if a restart occurs between them. Entry is delayed by one extra bar and fires at the right size (50%) eventually — no safety issue, but timing is affected every time the container restarts mid-pending. **Fix (optional, low priority):** Persist `_rsi_ext_pending_dir` to a short-TTL Redis key (`rsi_ext_pending:{sym}`, TTL 120s) on set; restore in `_evaluate_symbol` on first call. --- ### RISK (carried from 2026-05-31): `TV_BXT_NLV_PCT = 50.0` sizing bypass absent from `strategy_decisions.md` **File:** `bot.py:6885–6898` ```python _tv_pct = float(os.environ.get("TV_BXT_NLV_PCT", "50.0")) ``` $25K/trade on a $50K account; bypasses MAX_POSITION_PCT, ATR-risk sizing, all conviction multipliers. No hypothesis/backtest/validation record exists. --- ### RISK (carried from 2026-05-31): `config.json:max_concurrent_positions: 50` disables concurrent-position gate **File:** `config.json:24`; `core/risk.py:114` Gate fires only when ≥50 positions are open. With an 8-symbol watchlist, gate never fires on real money. --- ### RISK (carried from 2026-05-31): `config.json:risk_per_trade_pct: 2.0` bypassed by env-var sizing path **File:** `config.json:7`; `bot.py:6179–6210` Live sizing reads `RISK_ATR_BUDGET`/`MAX_RISK_PCT` env vars; `config.json` value is backtester-only. Three inconsistent values confuse operators. --- ### RISK (carried from 2026-05-31): Oracle `HEALTH_PORT` and bot `DASH_PORT` both default to `8083` **Files:** `oracle/config.py:56`; `bot.py:184` Silent dashboard outage if oracle wins the host-8083 binding on a compose-with-ports deploy. --- ### RISK (carried from 2026-05-31): `config.json:atr_stop_multiplier` (1.5) diverges from code fallback (2.0) **Files:** `config.json:8`; `bot.py:750` Removing or misspelling the key silently widens stops 33% with no log. --- ### RISK (carried from 2026-05-31): `pages.py:1448` hardcodes `"TradingView webhook (SNDK only)"` model label Label goes stale immediately if a second symbol is added to `TV_BXT_SYMBOLS`. --- ### RISK (carried from 2026-05-31): `sentiment.cashtags` seed mismatches trading `symbols` **File:** `config.json:2,37` `META`, `TSLA`, `PLTR` traded but absent from cashtag seed — sentiment bypassed ~15 min post-startup. --- ### RISK (carried from 2026-05-31): ~40+ undocumented feature-flag env vars; no `.env.example` `TV_BXT_DISABLE_ALGO_STOPS` (a stop-loss kill switch) is particularly dangerous if accidentally left set on a fresh reprovision. --- ## OK (checked, working or recently fixed) - **Time gate order (overnight gap):** `in_overnight_gap = 3*60+50 <= t_min < 4*60` correctly maps 03:50–04:00 ET to 00:50–01:00 PT. ✓ - **Close-hour gate (rsi_extreme):** `if in_close_hour and trigger_type != "rsi_extreme":` correctly blocks BXt/div/pullback 15:00–16:00 ET. Size scale 0.1× applied via `LAST_HOUR_SIZE_SCALE`. ✓ - **Friday entry cutoff (`_evaluate_symbol`):** `weekday() == 4 and t_min >= 15*60` at `bot.py:5521`. ✓ - **Friday entry cutoff (`_check_pullback`):** same condition at `bot.py:6361` cancels ALL pending including rsi_extreme (2026-05-19 P1 fix). ✓ - **RVOL blowoff ceiling:** Code at `bot.py:6027` correctly uses `> 5.0x` (tuned 2026-04-15). `CLAUDE.md` RVOL table correctly shows `> 5.0x`. ✓ - **RVOL fail-closed:** `rv is None` blocks entry at `bot.py:5979–5981`. ✓ - **RVOL floor exemption for rsi_extreme:** `not _is_rsi_extreme` guard on the 0.8× floor correctly lets through low-RV exhaustion setups. ✓ - **`total_scale` applied in `_open_trade`:** `bot.py:6860–6862` applies the composite multiplier to shares. ✓ - **Event-gate + block-profile apply to `total_scale`:** `bot.py:5935–5936` and `bot.py:6201–6202` — the 2026-05-03 sizing bug #2 fix is present. ✓ - **`signal_meta` None guard in `_open_trade`:** `bot.py:7002–7003` (`bd256c6` fix). ✓ - **`_pending_entries` atomic swap:** `bot.py:5856` uses tuple assignment to avoid race condition (2026-04-28 P0 fix). ✓ - **BXt RSI zone filter (dev):** `BXT_RSI_LONG_MAX=40` / `BXT_RSI_SHORT_MIN=60` applied AFTER time gates; no time-gate interaction issue. ✓ - **Off-hours rsi_extreme toggle:** `_is_rsi_extreme_offhours_enabled()` returns `False` when Redis key absent (fail-safe). ✓ - **`micro_scale` folded into `total_scale`:** `bot.py:5760` includes `micro_scale` in the product. ✓ - **Friday EOD flat sweep (3:50 PM ET):** `bot.py:4334–4347` closes ALL positions with no exemption. ✓ - **RSI extreme `startswith('RSI extreme')` fix:** EOD trim correctly exempts only `'RSI extreme'` prefix, not `'RSI divergence'` (2026-04-15 fix). ✓ - **`REVERSE_TIME_DECAY` morning ramp logic:** Computes `_frac` correctly at `bot.py:5754`; RTH-only guard (`_in_rth`) prevents off-hours application. ✓ - **`BXT_USE_15M_BIAS` counter-trend hard block:** `bot.py:5661–5672` blocks counter-trend BXt unless flag set to false (dev A/B). ✓ - **Single-model lock for BXt1m:** `bot.py:5892–5894` drops all original-model candidates when BXt1m allowlist non-empty. ✓