All arms net 10bps on rule toggles; split 2016-01-01; sqrt(252). PASS bar: OOS Sharpe >= 0.50 AND boot p < 0.05 AND residual Sharpe vs market > 0.
Rule Status Ret Sharpe MaxDD CorrMkt ResidSh boot-p
R2 breadth-capitulation ok -0.064213 -0.057418 -0.149316 0.382007 -0.429085 0.5708 R4 witching-week ok -0.193621 -0.203206 -0.392449 0.464168 -0.694632 0.7454 R5 vol-cone-fade ok -0.073786 -0.004003 -0.298011 0.652228 -0.769125 0.5040 R6 gap-drive ok -0.815709 -0.521686 -0.887938 0.330842 -0.925719 0.9502 R7 stressed-payer-short ok -0.876675 -1.057576 -0.888868 -0.999948 0.591506 0.9996 R1 pre-earnings-drift ok 2.612692 0.642339 -0.481853 0.554179 0.063531 0.0204 R3 window-dressing ok -0.158406 -0.356696 NaN NaN NaN 0.8794
ALL DONE
Reviewer note
-
Metrics above are OOS (split 2016-01-01), net of 10bps-on-toggle cost, annualized 252. R7 is corrMkt -1.0 (short EW basket) -> pure -beta, REJECT. R5/R4/R6 resid negative -> REJECT (vol-spike fade and witching-week longs are beta, gap-drive dies with prior-day volume filter; initial R6 pass of +58k% was a same-day volume leak, now fixed).
-
R1 (pre-earnings-drift) is the only rule meeting the bar (OOS SR 0.64, p 0.02, resid +0.06) but resid is fragile -> CANDIDATE for a placebo check (random trading days matching window frequency) + an IS/OOS per-window breakdown before any promotion. No placebo check run yet; do NOT sleeve.
-
R3 window-dressing REJECT (OOS sr -0.36).
-
R1 placebo FAILED: shuffled windows earn median +356.8% vs R1 +399.4% (p=0.42) -> window timing is NOT the edge; the apparent pre-earnings drift is unspecific beta. Parked, no sleeve.