Consensus capitulation (stacking triggers in the family) · Null results

The result

Code: research/consensus_capitulation.py Report: research/consensus_capitulation_report.json

For each (symbol, day) inside a drawdown ≥ 30 %, count how many distinct triggers from {death_cross + vol, bullish_engulf, gap_breakaway/runaway/exhaustion/island_down, bear_flag} fired in the trailing 10 trading days. Forward returns are monotonic in trigger count:

Trigger count × OOS 2020+	n	40d edge	60d edge	hit @ 40d
triggers = 1	6,818	+2.40 %	+4.24 %	58 %
triggers = 2	2,463	+5.39 %	+11.27 %	62 %
triggers ≥ 3 (OOS 2020+)	1,401	+9.05 %	+23.22 %	70 %

The 3+ subset is the highest-conviction sub-cohort in the repo at the 1-3 month horizon. n=1,401 is plenty for productization. Caveat: these are by construction names where lots of bad-news triggers fired on top of each other — single-name sizing risk is real (some names are correctly diagnosed as melt-downs and don't recover).

At a glance

Sample	1,401
Validator	research/consensus_capitulation.py
Verdict	Rejected

Every result here is reproducible from the script named above. Reports are in the repository.

Other rejected theses

All 13+ →