LAB #3670 HARMFUL: LEDGER BENCH: Apply filter: GBPUSD LONG @ NY_OPEN with 6 conf — improves win rate from 32.4% to 66.7 — E8 Intelligence Research

IDEA: Adjustment made to the live model — verify against the backtests: Apply filter: GBPUSD LONG @ NY_OPEN with 6 conf — improves win rate from 32.4% to 66.7% (+34.2pp, n=9) SAME-WINDOW EFFECT: Over the current live window, the filter would have blocked the two GBPUSD SHORT trades (rizzy5m and rizzy7m) since they are opposite direction to the proposed LONG filter. The rizzy5m GBPUSD SHORT closed as a WIN, so removing it would reduce win count by 1; the rizzy7m GBPUSD SHORT is still open, so its outcome is unknown. Net effect on the 10-trade window: win rate drops from 3/8 closed (37.5%) to 2/7 closed (28.6%), and PnL would be slightly negative if the open rizzy7m trade later loses — the filter would have hurt this specific window, not helped it. OVER-TIME PROJECTION: The claimed +34.2pp win-rate improvement is based on n=9, which is statistically meaningless — the 95% confidence interval spans roughly ±30pp, so the true effect could be anywhere from +4pp to +64pp. Over 3–6 months, t Author: Andrew Stewart Caldin, Independent Researcher, UK. Part of the E8 Intelligence Research series. Platform: e8intelligence.com