/parabolic-short-trade-planner

Intraday Trigger Playbook

Status (v0.5 — implemented): this reference describes the trigger semantics that monitor_intraday_trigger.py evaluates as a one-shot FSM. Phase 2 plans (generate_pre_market_plan.py) emit entry_hint / stop_hint formula strings that map to these triggers; Phase 3 reads bars (Alpaca live or fixture), walks the FSM, and writes intraday_monitor JSON with concrete entry_actual / stop_actual / shares_actual.

FSM contracts implemented in Phase 3:

• triggered is not terminal — post-trigger bars are still evaluated for invalidation predicates so a reclaim flips the plan to invalidated.
• invalidated is terminal/absorbing — no further transitions for that plan_id on that session date.
• First Red same-bar tie-break: when a single bar both takes out red_high (would invalidate) AND prints below red_low (would trigger), invalidation wins. Short-side conservatism.
• ORL post-trigger invalidation requires close > orl_low AND close > current_vwap (BOTH reclaimed, not just one).
• Idempotency: every Phase 3 run replays the full session bars; the FSM is a pure function of (plan, bars, atr_14). prior_state is consulted only for display continuity / notification diffs.
• Bar timing (v0.5d): ts_et on each bar is the bar-open instant (matches Alpaca wire). A bar with ts_et = T covers [T, T+5min) and is confirmed at T+5min. The Phase 3 adapters only return confirmed bars (bar_open + 5min <= until_et); the FSM never sees an in-progress current bar.
• ORL anchoring: the Opening Range bar must be the 09:30 ET bar. When Alpaca skips that interval (no trades / halt at the open), the ORL evaluator emits evaluation_status="skipped" + skip_reason="opening_range_bar_unavailable" rather than anchoring on a later bar.

5-min Opening Range Low (ORL) break

plan_id template:   <TICKER>-<YYYYMMDD>-ORL5
trigger_type:       orl_5min_break
condition (ja):     5min ORL を出来高 1.2x 以上で下抜け
entry_hint:         5min_orl_low - 0.05
stop_hint:          session_HOD + 0.25 * ATR
structural_targets: dma_10, dma_20

Phase 3 evaluator:

1. Wait for the first 5-minute bar to close. Mark its high (ORH) and low (ORL).
2. After 9:35 ET, watch every 5-minute close. If a bar closes below ORL AND its volume ≥ 1.2× the ORL bar's volume, fire the trigger.
3. Stop is session_HOD + stop_buffer_atr * ATR(14).
4. Invalidates if a subsequent 5-minute close prints back above ORL AND VWAP (sign of failed breakdown).

First Red 5-minute candle

plan_id template:   <TICKER>-<YYYYMMDD>-FR5
trigger_type:       first_red_5min
condition (ja):     寄付後最初の赤 5min の安値割れ
entry_hint:         first_red_5min_low - 0.05
stop_hint:          first_red_5min_high

Phase 3 evaluator:

1. Track every 5-minute bar from 9:30 ET. Mark the first bar where close < open (a red candle).
2. After that bar closes, fire the trigger when a later 5-minute bar prints below the red candle's low.
3. Stop is the red candle's high.
4. Invalidates if any subsequent 5-minute bar takes out the red candle's high before the trigger fires.

VWAP fail

plan_id template:   <TICKER>-<YYYYMMDD>-VWF
trigger_type:       vwap_fail
condition (ja):     First crack 後 VWAP retest で 5min 終値拒否 + lower-high 下抜け
entry_hint:         lower_high_low - 0.05
stop_hint:          vwap_reclaim_5min_close

Phase 3 evaluator (FSM):

State machine has six states: armed → first_crack_seen → vwap_retest_seen → rejection_confirmed → triggered → invalidated

1. armed: market open, watching for first VWAP loss.
2. first_crack_seen: 5-minute close prints below VWAP for the first time, AND price comes from session HOD (filters open-print VWAP noise).
3. vwap_retest_seen: subsequent bar closes back at or above VWAP (the "retest").
4. rejection_confirmed: next 5-minute bar prints a lower high than the retest bar AND closes back below VWAP.
5. triggered: fire on the break of the rejection bar's low.
6. invalidated: any 5-minute close back above VWAP after triggered OR a clean uptrend line break before triggered.

Common evaluator concerns

• Time zone: all timestamps America/New_York.
• Halt handling: bars during halt are skipped; the FSM resumes from its last state when trading resumes.
• Bar close vs touch: triggers fire on bar close, not intra-bar. Reduces wick noise.
• No re-entry: once a plan is triggered or invalidated, the FSM does not re-arm for the day.