LAB #2535 NEUTRAL: LEDGER BENCH: Apply filter: Session = OFF_SESSION — improves win rate from 31.0% to 44.4% (+13.5pp, — E8 Intelligence Research

IDEA: Adjustment made to the live model — verify against the backtests: Apply filter: Session = OFF_SESSION — improves win rate from 31.0% to 44.4% (+13.5pp, n=99) SAME-WINDOW EFFECT: Applying the OFF_SESSION filter would have blocked the 11:35, 10:30, 10:10, and 11:20 entries (all within the filtered session), removing 4 trades from the 10 shown. Of those removed, 3 were wins (GBPUSD 11:35, GBPUSD 10:10, USDCAD 11:20) and 1 was a loss (XAUUSD 11:35 on 08-04). This would drop the live win rate from 6/10 (60%) to 3/6 (50%) — a *decrease* in this window, contradicting the backtest improvement. PnL delta would be negative, as you'd forfeit three winners to avoid one loser. OVER-TIME PROJECTION: The backtest gain (+13.5pp on n=99) is statistically weak (p≈0.08, not significant at 95% CI). Over 3-6 months, the filter may reduce trade frequency by ~30-40%, cutting gross PnL even if win rate improves, because you're removing a large chunk of the sample. The main risk is overfitting to a spe Author: Andrew Stewart Caldin, Independent Researcher, UK. Part of the E8 Intelligence Research series. Platform: e8intelligence.com