Forward-Shadow Test · halfsize overlay vs V5

Frozen 2026-06-01 · last bar 2026-08-09 23:45 · updated 2026-08-10 06:02 UTC

70days elapsed

27baseline trades

27overlay trades

⚠ Early sample — 27 trades is far too few to conclude anything. This page accrues daily; the signal only becomes meaningful after the overlay has resized a few dozen mid-vol-band entries.

Head-to-head (trades entered ≥ freeze)

metric	V5 baseline	halfsize overlay	Δ
trades	27	27	0
win %	59.3%	59.3%	—
net PnL	6854.32	6864.44	+10.12
profit factor	1.443	1.443	0.0
mean R	0.2025	0.2025	0.0

What this is

Post-freeze bars are the virgin out-of-sample the Study-2 hypothesis never had — the dead-zone band [0.33, 0.66) was chosen using IS+OOS, so only data from 2026-06-01 onward is genuinely clean. Both arms run the identical V5 engine over the same fresh bars; the overlay half-sizes mid-vol-band entries. If the overlay's edge is real, its mean-R and PF will pull ahead as the sample grows. If they stay tied or the overlay trails, the Study-2 result was in-sample noise. Updated daily by cron.