A 135,356-trade historical backtest showed asymmetric (1−P)/P payoffs in the cheap band — the upside dwarfed the cost of being wrong even at modest hit rates, while buys above 0.60 were net-negative even at high hit rates. Live data narrowed the band further: 0.30–0.40 died in wave 29 (n=20, 17.6% hit, −16pp), 0.05–0.20 in wave 31. Live band is now 0.20–0.30.

We fire when a wallet drops $25K+ in that band and hold to actual market resolution. No 24h directional drift games. The running hit rate is the panel directly below this one — early sample, treat accordingly.