MODEL ADOPTION #3905: REJECT — LEDGER BENCH: Apply filter: Session = NY_OPEN — improves win rate from 32.5% to 39.0% (+6. — E8 Intelligence Research

DECISION: REJECT ELEMENT: No element survives — the NY_OPEN filter is a session-label proxy, not a geometric timing edge. E8-node/Gods-Clock rhythm logic already encodes time-as-structure; adding a wall-clock session filter would double-count time and risk misaligning with the model's internal phase boundaries. EXPECTED IMPACT: Win-rate gain is likely real but not additive — it would come at the cost of ~60–70% trade frequency, reducing compounding and increasing variance. The +6.5pp is within noise for n=777 when split across sessions (confidence interval ±4–5pp at best). Net PnL improvement is marginal and not robust to regime shifts. RISK: Overfitting to NY_OPEN's current volatility clustering. During low-vol regimes or news spikes (NFP, FOMC), the filter would either starve the model of trades or admit false signals that happen to fall in the window. Also, the 11:05 win barely squeaks in — a 5-minute shift would flip the result. TEST: Before any re-adoption, run a 12-month, Author: Andrew Stewart Caldin, Independent Researcher, UK. Part of the E8 Intelligence Research series. Platform: e8intelligence.com