Gamma Shift v6

WHAT CHANGED FROM v5: v5 used total positive gamma tilt (all participants averaged). v6 weights them: 50% Firm + 30% MM + 20% Customer This doubled the predictive power (IC 0.182 vs 0.094 OOS). WHY PARTICIPANTS MATTER: • Firm tilt = smart money positioning. When firms have positive gamma above spot, they're positioned for upside. Most predictive individual participant. • MM tilt = dealer hedging. Useful in the blend but CONTRARIAN when extreme (MM >65% = bearish, not bullish — 25% WR when MMs pile above spot). • Customer tilt = retail/institutional flow. Confirms direction, not contrarian. LONG SIGNALS: 🟢🟢 VERY HIGH (71% WR): weighted tilt >65% + firm neg gamma above + RVOL <1.0 🟢 HIGH (71% WR): weighted tilt >55% + firm neg gamma above + RVOL <1.26 🟡 MEDIUM (62% WR): weighted tilt >55% + RVOL <1.26 ⚪ LOW (55% WR): weighted tilt >50% + RVOL <1.26 SHORT SIGNALS: ⚠️ CONFLICT (59% short): weighted tilt >55% BUT firm neg gamma BELOW spot The gamma wall looks bullish but firms are positioned for downside. Trap! 🔴 BEARISH (65% short): weighted neg gamma (50F/50MM) <30% above spot Both firms and MMs have downside fuel loaded below spot. KILL CONDITIONS: 🚫 RVOL ≥ 1.26 (top quartile) → ALL gamma signals die. No edge in high volume. 🚫 Outside 6:30-8:00 PT → signal window only works in the morning. KEY FINDINGS FROM BACKTEST (343 clean days, IS/OOS validated): • Total tilt alone: 68.8% WR at >55% — still works but participant blend is better • Weighted blend: 71.4% WR at >55% — 2x the IC of total • RVOL is the #1 kill condition — high RVOL + any tilt = 17% WR (disaster) • Firm neg gamma confirms or denies: pos wall + firm neg above = 67.5%, conflict = 40.7% • v6 fires ~3x more often than v5 (12.5% vs 4.4% of days)