Pocket pivot (O'Neil / Morales)

Thesis (rejected): Inside a base / Stage-2 uptrend, an up-day whose volume exceeds the highest volume of any DOWN day in the prior 10 sessions is a stealth-accumulation footprint that precedes the visible breakout.

Code: research/pocket_pivot.py Report: research/pocket_pivot_report.json (n=134,049)

OOS-2020+ doesn't help (20d edge +0.09 %). Adding a Stage-2 trend gate also fails to rescue it: Stage-2 vs non-Stage-2 buckets are indistinguishable on every horizon. n is enormous (134k events), so this is not a statistical-power problem — the within-base single-day volume signature simply does not predict forward returns once the universe is broad enough to remove selection bias.

Verdict: Reject. The "stealth accumulation" intuition is present in O'Neil's hand-picked anecdotes but does not generalise.