MODEL ADOPTION #897: REJECT — LEDGER BENCH: Apply filter: Phase distance TIGHT (<=2deg) — improves win rate from 31.6% t — E8 Intelligence Research

Abstract

DECISION: REJECT ELEMENT: The phase distance filter (≤2°) is not adoptable because the backtest (n=391) is too small to distinguish signal from noise, and the live shadow-test proposed is a 30-day forward test, not a replay-bench against historical E8-node/Gods-Clock data. The 3.7pp win-rate gain is fragile — it could vanish in a different market regime (e.g., trending vs. choppy) because phase distance tightness is regime-dependent. The "promising" verdict at 68% confidence is insufficient for a live model that already has validated FADE polarity edge at scale. EXPECTED IMPACT: If adopted, win rate might temporarily rise ~2-3pp, but trade frequency would drop ~15-20%, likely reducing absolute monthly profit by 5-10% due to missed high-reward setups at wider phase distances. The net effect is negative or neutral. RISK: Overfitting to recent phase alignment conditions. The filter would reject trades that later prove profitable in wider-phase regimes, degrading model adaptability. Als Author: Andrew Stewart Caldin, Independent Researcher, UK. Part of the E8 Intelligence Research series. Platform: e8intelligence.com