MODEL ADOPTION #4967: REJECT — LEDGER BENCH: Apply filter: XAUUSD LONG — improves win rate from 31.7% to 54.1% (+22.4pp, — E8 Intelligence Research

DECISION: REJECT ELEMENT: Nothing survives as a standalone rule. The only defensible element is a *conditional* version — "XAUUSD LONG only when ATR-percentile ≥ 60 AND higher-timeframe trend state = UP" — but that condition is not what Lab #4967 tested, so it cannot be adopted on this evidence. The raw rule "XAUUSD LONG" is a direction+instrument filter with no causal mechanism, and it is incompatible with E8/Gods-Clock geometry as stated because it overrides the rhythm logic on one instrument without a regime gate. EXPECTED IMPACT: Honest estimate: ~0 to +1pp on live win rate, not +22.4pp. The +22.4pp is almost certainly regime-fitting: n=85 with a 22pp swing on a single instrument is the classic signature of a filter that selected a trending-gold window. In the live window it removes exactly one loser (the 2026-09-10 XAUUSD SHORT) — that is a 1-observation effect, statistically indistinguishable from noise (binomial p ≈ 0.5). Profit impact is likely negative-to-flat because the fi Author: Andrew Stewart Caldin, Independent Researcher, UK. Part of the E8 Intelligence Research series. Platform: e8intelligence.com