Sydnee nightly — exit_flow audit — 2026-05-27 P0 findings: 0 P1 findings: 2 Risks: 3 - Area: Exit flow (how and when open positions are closed) - Bugs found (P0 / P1): 0 / 2 - Risks noted: 3 Branch: `dev` (7147326). Cross-checked against `git log --oneline -30` and `docs/strategy_decisions.md`. Both P1 bugs are **carryovers from 2026-05-26 that were not fixed in today's commits** (7147326, 94aeea3, 7f1fd63, f631fd2, ee2a66d, 643b2a4, 3d8bdda, 2928bd1, c51d818, 8535f8e, d4aa901). One new RISK found in the dynamic TV-BXt symbol set. --- Full report (dev branch): https://github.com/kanex1/sydnee.signals/blob/dev/docs/audit_2026-05-27_exit_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 2026-05-27 — EXIT FLOW ## Summary - Area: Exit flow (how and when open positions are closed) - Bugs found (P0 / P1): 0 / 2 - Risks noted: 3 Branch: `dev` (7147326). Cross-checked against `git log --oneline -30` and `docs/strategy_decisions.md`. Both P1 bugs are **carryovers from 2026-05-26 that were not fixed in today's commits** (7147326, 94aeea3, 7f1fd63, f631fd2, ee2a66d, 643b2a4, 3d8bdda, 2928bd1, c51d818, 8535f8e, d4aa901). One new RISK found in the dynamic TV-BXt symbol set. --- ## Findings ### BUG [P1]: OBV Rule F permanently disabled after in-loss skip (carryover 2026-05-26) **File:** `bot.py:4455–4477` **Evidence:** ```python if _profit_guard and profit_per_share < 0: _dir_lbl_ex = "LONG" if t.direction == 1 else "SHORT" if _loss_exit: hit = True reason = (...) elif not getattr(t, '_obv_be_applied', False): self._log_activity(...) # logs once, then silent t._obv_be_applied = True # ← line 4475: flag consumed in in-loss path t._obv_be_applied_at = datetime.now(ET).isoformat() self._update_trade_field(t.trade_id, obv_be_applied=True, ...) else: # In-profit path — tighten stop ... t._obv_be_applied = True # ← line 4481: correct (stop moved) ``` The outer guard at `bot.py:4409` is `and not getattr(t, "_obv_be_applied", False)`. Once the flag is True, Rule F never fires again. The in-loss path (profit guard) sets the flag on line 4475 **without moving the stop**, consuming the one-shot protection permanently. **Impact:** Trades that go underwater before OBV turns adverse, then recover to profit later, lose their Rule F stop-tighten. The wide entry stop defends until hit or another exit fires. The −$774→−$217 avg damage reduction (bxt_reversal backtest) only applies to trades that are in-profit when Rule F first fires. **Fix:** Move `t._obv_be_applied = True` inside the `else:` branch so it's only set when the stop actually moved: ```python if _profit_guard and profit_per_share < 0: if _loss_exit: hit = True ... elif not getattr(t, '_obv_be_applied', False): self._log_activity(...) # no-op this bar only # Do NOT set _obv_be_applied — let Rule F retry next bar when profitable else: ... t._obv_be_applied = True # only mark applied when stop actually moved ``` Note: `OBV_LOSS_EXIT` (spot-exit) path can keep the flag set because the trade is being closed — that's already correct (hit=True). --- ### BUG [P1]: TRIM2 `_rsi_force_trim2` flag never cleared for 1-share parents (carryover 2026-05-26) **File:** `bot.py:4871–4905` **Evidence:** ```python if not hit and is_trimmed and not is_trimmed_2 and (_trim2_hit or _force_trim2): trim = True # ← set unconditionally (line 4872) trim2_shares = t.shares // 2 if trim2_shares > 0: ... t._rsi_force_trim2 = False # ← only cleared when trim2_shares > 0 self._update_trade_field(t.trade_id, trimmed_2=True, force_trim2_pending=False, ...) # No else branch ``` When TRIM1 reduces position to 1 share, `trim2_shares = 1 // 2 = 0`. The `_rsi_force_trim2` and `trimmed_2` flags are inside the `if trim2_shares > 0:` guard — they are never updated. On every subsequent bar, `_force_trim2` returns True, `trim = True` is set, and the normal activity log at `bot.py:5034` is suppressed (`elif not trim:`). **Impact:** Per-bar CPU waste and silent activity log suppression for every bar until the position exits via stop, EOD flat, or target. The `force_trim2_pending` column stays True in DB, surviving restarts. No incorrect close is generated — the no-op is harmless to P&L but obscures the trade's live status in logs. **Fix:** Add an `else` branch after the `if trim2_shares > 0:` block: ```python else: # 0 shares — can't trim. Mark done so we don't retry every bar. t._trimmed_2 = True t._rsi_force_trim2 = False self._update_trade_field(t.trade_id, trimmed_2=True, force_trim2_pending=False, force_trim_reason=None) ``` --- ### RISK: Dynamic TV-BXt symbol position orphaned during Redis failure **File:** `bot.py:3571–3589` (`_tv_bxt_active_syms`) **Evidence:** ```python def _tv_bxt_active_syms(self) -> set: syms = self._tv_bxt_env_syms() # env symbols only (e.g. SNDK) try: r = getattr(self, "_regime_redis", None) if r is not None: for v in (r.smembers("tv_bxt_dynamic_syms") or []): syms.add(v.upper()) except Exception: pass # ← Redis failure silently dropped return syms ``` `_eval_tv_bxt` (line 3828) iterates over `_tv_bxt_active_syms()` and calls `_eval_tv_bxt_sym(sym, win)`. The overnight flat (`win == "flat"` → force-close), OBV trail stop, and Friday EOD flat all live inside `_eval_tv_bxt_sym`. If Redis is temporarily unavailable, the try/except swallows the error and returns only env symbols. **Scenario that breaks:** Model is "tv" (SNDK in env → `_bxt_model()` returns "tv"). A TV-BXt position is open for MU (added to `tv_bxt_dynamic_syms` via webhook). Redis goes down. `_tv_bxt_active_syms()` returns only `{SNDK}`. `_eval_tv_bxt_sym(MU, ...)` is never called. The model-inactive path (line 3822–3826) doesn't fire because the model is still "tv". MU's position: - Never gets its overnight flat checked - Never gets its OBV trail stop checked - Never gets its Friday EOD flat checked The position becomes an orphan until Redis recovers. **Mitigation:** Add a safety sweep in `_eval_tv_bxt` that closes any open TV-BXt position whose symbol is not in the current active set: ```python active = self._tv_bxt_active_syms() for t in list(self.open_trades()): if self._is_tv_bxt_trade(t) and t.symbol not in active: # Symbol dropped from active set (Redis failure or TTL expiry) # Force-close to prevent orphan. win_sym = self._bxt1m_window(datetime.now(ET)) if win_sym == "flat": self._close_trade(t, "TV-BXt FLAT — symbol dropped from active set") self._tv_bxt_signal_clear(t.symbol) ``` Or simpler: log a WARN and let the next successful Redis read recover naturally (acceptable given the 7-day TTL makes overnight TTL expiry essentially impossible; real risk is only transient Redis failure). --- ### RISK: Stale "keep overnight" comments in Friday EOD progressive-trim block (carryover 2026-05-26) **File:** `bot.py:4078, 4082, 4091` **Evidence:** ```python # Skip RSI extreme entries from after 12:00 PM PT — keep those intact. # ← wrong # weekend. Only rsi_extreme is meant to keep overnight. # ← wrong continue # skip — late RSI extreme, keep overnight # ← wrong ``` The overnight-hold exemption for late RSI extreme entries was removed in commit `52f1bb1` (2026-04-17). The final flat sweep at `bot.py:4139–4158` now closes ALL positions, no exemptions. The progressive-trim exemption (3:00–3:50 PM ET window) is scoped to **trimming only** — it prevents a fresh entry from being immediately trimmed, but the position is still closed at 3:50 PM. The comments say "keep overnight" which is incorrect. **Impact:** No runtime bug — positions are not actually held overnight. The misleading comments risk a future developer re-introducing the overnight exemption assuming it was an intentional ongoing behavior. **Mitigation:** Update comments at lines 4078, 4082, 4091 to: "Skip late RSI extreme entries from progressive trimming only — position still closes at 3:50 PM ET flat sweep." --- ### RISK: Stale trim3 DB-column comment (carryover 2026-05-26) **File:** `bot.py:7787` **Evidence:** ```python elif trade.trade_id.endswith("_trim3"): pass # no DB column for trim3 yet — state tracked in-memory ``` The `trimmed_3` column WAS added in the 2026-04-15 persistence fixes (commit `a6be962`, `strategy_decisions.md` line 344). It is written at EXIT 3 (`bot.py:4835–4836`) before the trim child is queued. The `pass` here is correct (parent already updated), but the comment is false. **Impact:** No runtime impact. A developer reading this comment might attempt to "fix" it by adding the update here, causing a benign duplicate write. **Mitigation:** Update comment to: `# trimmed_3 already set via _update_trade_field in EXIT 3 (bot.py:4835) — no second update needed`. --- ## OK (checked, working or recently fixed): - **`_tv_bxt_signal_clear` on all exit paths** — verified: called on overnight flat (line 3844), OBV trail stop (line 3874), Friday EOD flat (line 3879), model-switch-into-tv signal purge (line 13226), manual close (line 13019), IBKR-side reconcile close (line 2390). - **Dynamic symbol single-model lock** — `_bxt_active_syms()` correctly calls `_tv_bxt_active_syms()` in "tv" mode (line 3769), so the single-model suppression covers both env and dynamic symbols. - **Switch-into-TV signal clear** — now clears both env and dynamic symbol signals via `_tv_bxt_active_syms()` (line 13226), correctly superseding the old env-only clear. - **"hold" window removal for TV-BXt (f631fd2)** — entry during 09:15–09:45 ET and 15:45–16:15 ET transition windows is now allowed. Friday EOD flat fires on the next tick for positions opened during 15:45 ET window on Friday. OBV trail stop and outsideRth=True order routing remain active. No orphan risk from this change. - **IP allowlist (`_tv_bxt_allowed_ips`)** — correctly defaults to TV's 4 official relay IPs; empty env string disables check (secret-only). Not an exit flow concern. - **Auto-add to dynamic set (7147326)** — webhook handler calls `r.sadd("tv_bxt_dynamic_syms", sym)` + `r.expire(..., 7*86400)` on each valid webhook. 7-day TTL means set survives active webhook streams indefinitely. Exit flow for dynamically-added symbols is correct when Redis is healthy (see RISK above for failure mode). - **FLAT_EXIT parent `exit_reason=""` in DB** — unchanged from yesterday's RISK finding (bot.py:1315). No new code touched this path. - **OBV trail Redis restore wired (498275d)** — `_obv_trail_saved` tuples persisted and restored from Redis at startup. Verified in `_load_trades_from_db` (bot.py:1129–1140 area, including 2026-05-26 additions). - **TV-BXt exempt from `_evaluate_symbol`** — confirmed (bot.py:4050–4052): `strategy_reason.startswith("BXT1m")` / `startswith("TV-BXt")` check prevents normal stop/trim/EOD logic from running on these positions (handled exclusively by `_eval_bxt_1m_naked` / `_eval_tv_bxt_sym`). - **All items verified OK in 2026-05-26 audit** — re-confirmed unchanged.