MODEL ADOPTION #1935: REJECT — LEDGER BENCH: Apply filter: Session = OFF_SESSION — improves win rate from 31.9% to 44.9% — E8 Intelligence Research

DECISION: REJECT ELEMENT: No element survives — the OFF_SESSION filter is not a concrete, testable edge compatible with E8 geometry timing; it is a time-window heuristic with no causal link to the rhythm logic. EXPECTED IMPACT: Win-rate lift is likely 3–5pp at best, not 13pp — the backtest n=98 is too small, and the live window showed only 5.6pp improvement. Profit impact is negligible because OFF_SESSION trades are fewer and lower-conviction; the filter would also cut valid wkjudas entries that already outperform. RISK: Overfitting to a 98-trade sample; regime shift where OFF_SESSION becomes high-volatility (news/crypto bleed); filter interacts poorly with rizzy5m vs wkjudas — could remove the few high-R winners that carry the model. TEST: Before any adoption, run a 90-day shadow replay on the live engine: tag every signal with SESSION=ON/OFF, but do NOT filter. Require ≥150 OFF_SESSION trades. Pass condition: OFF_SESSION win rate ≥40% AND ON_SESSION win rate ≤35% AND OFF_SESS Author: Andrew Stewart Caldin, Independent Researcher, UK. Part of the E8 Intelligence Research series. Platform: e8intelligence.com