Asymmetric MOO→11:00 v2

Apr 26 16:55 · honest tech proxies (no sector LA) · splits filter · per-min walker · per-ticker/week cap · dual liquidity gate ($500K)

v2 fixes: sector dropped → 5 technical proxies (mcap/price/atr/prev5d/RVOL_pm) · splits filter via strategies/splits_loader.py (3,366 rows / 0.44%) · per-minute first-touch walker · 1 trade per ticker per ISO week · liquidity gate PM≥$500K AND RTH≥$500K (kept 30.7% of 757K events).

Compound growth $10K (realized backtest)

Capital cap = min(20% capital, 5% pm_dvol, $1K floor) · costs 14 bps slippage RT + $2/trade + 0.5 bps SHORT locate · 5y day0 cache stops at 10:00 → backtest = 24 mo (2024-04 to 2026-04).

Forward projection (blended IS/OOS, capacity-capped)

Y1 219% blend (30% IS 44% + 70% OOS 294%). Plateau ~$49.6M (5% of $198M daily taken-pm-dvol). OOS only 8 months — Y1 likely optimistic.

Top 30 edges (sorted by mean_test)

#	Name	Side	Stop	Target	N_test	WR_test	mean_test %	W_mean	L_mean	PF	W/L

Per-minute walker vs conservative — uplift

Edge	Side	Stop	Target	N	Cons mean %	Walk mean %	Uplift %

Walker delivers +25% avg uplift on stop+target edges (max +1100% on gap_ge10 × c5_le-10 SHORT). Edges without target unchanged (no 2nd event). Conservative v1 systematically under-counted target-first hits.

Earnings 5y deep-dive — 6 edges (weak as primary)

Name	Side	Stop	Target	N_test	WR	mean %	PF	W/L

Earnings = confirmation/size modifier, not primary filter. Best: AMC d-1 × gap 2-5% LONG stop -1 target +3, PF 2.93 N=36. (Consistent with prior research_earnings_audit_apr22.md).

Live deployment checklist

Создать отдельный модуль moo_1100_asym/ (не трогать moo_955/)
Грузить top-20-30 distinct edges из edges_v2.parquet
Pre-trade gates: PM dvol≥$500K live · splits check · per-ticker/week cap · max 4 concurrent positions
Position size = min(20% capital, 5% pm_dvol, 50% capital, $1K floor)
Entry MOO 09:30 · exit walker first-touch stop/target иначе принудительно 11:00
Halt при -25% drawdown от peak
Re-check live PM dvol на 09:25-09:29 перед MOO submit

Limitations

2y window только (5y day0 cache stops at 10:00 → earnings deep-dive ограничен 2y)
OOS только 8 месяцев → Y1 219% projection оптимистичный
Walker tie within same minute = conservative stop-first (5-10 bps downward bias)
SHORT-heavy basket (58 SHORT vs 23 LONG) — borrow rate в live execution