MODEL ADOPTION #2602: REJECT — LEDGER BENCH: Apply filter: XAUUSD LONG — improves win rate from 31.0% to 58.1% (+27.1pp, — E8 Intelligence Research

DECISION: REJECT ELEMENT: No element survives — the XAUUSD LONG-only filter is a regime bet, not a structural edge. E8 geometry timing is polarity-agnostic; forcing a directional bias violates its core rhythm logic. EXPECTED IMPACT: Win-rate gain is real in the current window (+27.1pp) but statistically fragile (n=74, CI ±11pp). Over 3–6 months, if gold trends down, the filter blocks all winning shorts and forces losing longs — net PnL likely negative vs. baseline. Frequency drop also reduces compounding. RISK: Regime shift to bearish gold → model sits out or takes only losing longs. Also, the filter corrupts E8's polarity symmetry, potentially degrading signal quality across all pairs via shared rhythm state. TEST: Must pass a 3-year quarterly regime-split backtest (bull/bear/sideways) showing LONG-only edge persists in bear phases with ≥30 signals per regime and positive R-multiple. Also require a 60-day shadow test on live engine comparing filtered vs. unfiltered on total Pn Author: Andrew Stewart Caldin, Independent Researcher, UK. Part of the E8 Intelligence Research series. Platform: e8intelligence.com