Bullish divergence × consensus — highest-conviction stack · Null results

The result

Code: research/div_consensus_intersect.py Report: research/div_consensus_intersect_report.json

For each bullish RSI/MACD divergence event, count how many other consensus triggers (death cross, gap_down, bullish_engulf, bear_flag) fired within ±10 trading days on the same name. Edge stacks monotonically with the consensus count:

rsi_div × consensus count × OOS 2020+	n	40d edge	hit @ 40d
no_consensus (0)	2,836	+6.9 %	71 %
1+ trigger nearby	2,047	+11.3 %	76 %
2+ triggers nearby	748	+16.7 %	84 %
3+ triggers nearby	304	+19.7 %	89 %
dd ≤ -30% + 3+ triggers	225	+21.6 %	90 %

n=225 is plenty for productization. 90 % of these events produce a positive 40-day return — the highest-conviction sub-cohort in the entire repo. Same story holds for macd_div (n=250, +19.8 % at 40 d, 84 % hit).

At a glance

Sample	225
Validator	research/div_consensus_intersect.py
Verdict	Rejected

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

Other rejected theses

All 13+ →