2026-11-01Backtest re-run with the 28 Oct look-ahead fix (entries now at the next open, because the signal is only known at the close). Stop 1%, target 2%, 5-day hold, trades / hit rate before -> after: reversal SPX 70/37.1% -> 70/40.0%, SPY 76/42.1% -> 74/41.9%, QQQ 100/49.0% -> 94/39.4%, IWM 99/35.4% -> 92/34.8%; exhaustion SPX 104/50.0% -> 96/63.5%, SPY 98/44.9% -> 93/53.8%, QQQ 35/45.7% -> 33/60.6%, IWM 8/50% -> 8/75%. The reversal results got weaker (QQQ by almost ten points); every reversal hit rate is at or below 42%. Exhaustion hit rates rose, but these are raw hit rates on one parameter set with no matched base rate, and IWM has 8 trades. The /backtest page computes these live, so it already shows the corrected figures. The volatility-regime (HMM) history chart is labelled in-sample: each past day is decoded with the whole 252-day window, including later days.
2026-10-30SPX Buy Zone trigger moved from 7,565.56 to 7,575.33 (SPY 754.06 to 755.11). Not an error in the arithmetic: a new Q.E. alert fired on 29 Oct and the zone is always measured from the most recent alert. The old zone was 2.36% below the 13 Oct alert's day 4-5 pivot high (7,748.42). The new alert's day 4-5 pivot has not printed yet, so the zone is anchored provisionally on its swing high (7,758.43, 28 Oct) and will move again when the pivot prints (4-5 Nov). The stored alert now records which anchor was used, and /brief prints each trigger's base level, rule and previous value.