Burst-chain options-exit grid

Question: Mirror the coil-breakout exits study for the new burst cohort. Which (spread, tenor, ruleset) combination maximises edge and Sharpe on the day-2 confirmation entry?

Code: research/burst_options_exits.py Report: research/burst_options_exits_report.json

Setup: Entry at close[T+1] (the day-2 confirmation). Filtered to second_burst == True from burst_continuation_report.json, optionable US universe, MIN_YEAR=2010. Grid: 5 spreads × 3 tenors (21/30/45 days) × 11 exit rules = 165 cells per (event, control) group on n=718 events / 3,511 controls. Mark-to-market is Black-Scholes with entry IV = HV30 * 1.30 held constant.

Top 5 by edge_mean (event mean − control mean):

Three observations:

1. Hold-to-expiry sweeps the leaderboard. Every top-15 cell by edge_mean is hold-to-expiry. Every TP-capped ruleset (tp100_sl50, tp75_sl50_t10, etc.) clips the right-tail and loses mean. Same pattern as coil-breakout exits: these cohorts are fat-right-tailed enough that early TPs sacrifice alpha.

2. 45-DTE > 30-DTE > 21-DTE. Despite the underlying drift being concentrated in the first 20 sessions, the longer tenor still wins — the extra theta room buys the position survival space during early choppiness. 21 DTE shows up only at the bottom of the table.

3. Tighter spread (1.0/1.10) > wider. The 1.0/1.10 spread costs more debit and caps at a lower max — but the cohort's modal winner is +5-15 % over 45 days, which fully consumes a 10-point spread. Wider spreads dilute the per-point return.

Comparison to coil exits: coil winner is also 1.0/1.10 hold-to- expiry, but at 60-DTE with event mean +57.9 % and edge +31 pp on n=241. Burst's edge magnitude is smaller (+14.2 pp) but n is 3x larger and the regime mix is more diverse. Sharpe comparable (0.22 vs 0.44 — coil is higher per-trade).

Risk-management note: The validation winner has NO stop. Live scanner spec carries a pct_of_risk_50 stop as a sanity check (same convention as coil), even though the grid said stop-free hold-to-expiry was strictly better. The stop is for operator comfort, not edge.

Verdict: Set the burst cohort's OPTIONS_SPECS entry to {kind: debit_call_spread, k_long: 1.00, k_short: 1.10, dte: 45, premium_frac: 0.45, stop_rule: pct_of_risk_50} and add a matching cohort_position_summary block. Done in web/scanner_charts.py:91-100,217.