LAB #4806 NEUTRAL: LEDGER BENCH: Apply filter: XAUUSD SHORT — improves win rate from 31.8% to 45.0% (+13.2pp, n=40) — E8 Intelligence Research

IDEA: Adjustment made to the live model — verify against the backtests: Apply filter: XAUUSD SHORT — improves win rate from 31.8% to 45.0% (+13.2pp, n=40) SAME-WINDOW EFFECT: The filter would have blocked the XAUUSD SHORT on 2026-09-10 — the one trade in the live window that closed as a LOSS. That's a single trade removed from a window of ~10 logged entries, so the realized delta is one avoided loss, not a +13.2pp win-rate shift; you cannot manufacture 13.2 percentage points from a sample where only one XAUUSD short exists. Net effect on this window: marginally positive PnL, statistically meaningless. OVER-TIME PROJECTION: The claimed +13.2pp rests on n=40, which is far too small to distinguish a real edge from noise — the 95% CI on a 31.8%→45.0% shift at n=40 spans roughly zero. Worse, the filter is symbol+direction specific (XAUUSD SHORT only), so it is almost certainly fitting a regime (recent gold uptrend punishing shorts) rather than a structural property. Over 3-6 months, if go Author: Andrew Stewart Caldin, Independent Researcher, UK. Part of the E8 Intelligence Research series. Platform: e8intelligence.com