- W1 EW top-10: OOS ret +1156.9%, SR 1.07, maxDD -36.2%, residSR +0.41, boot-p 0.000
- W2 invV63 top-10: OOS ret +758.0%, SR 1.00, maxDD -34.5%, residSR +0.29, boot-p 0.001
- W3 invV21 top-10: OOS ret +770.2%, SR 1.01, maxDD -34.5%, residSR +0.30, boot-p 0.001
101-name EqMom universe, 12-1 skip? no, raw L252 signal? used 12m pct change top-10, hold 21, next-open-ish, 10bps.
ALL DONE
Verdict: W2/W3 inverse-vol sizing cuts maxDD by only ~2pp (-34.5% vs -36.2%) at a ~0.07 SR cost -> REJECT on SR; no DD win worth the trade. EqMom EW sizing stands. The other free improvement channels (hysteresis, quality filters, universes, fundamentals overlay) have closed before; residual improvements live at book/overlay layer (already graduating C7/B5 via shadow).