MODEL ADOPTION #1945: REJECT — LEDGER BENCH: Apply filter: Quarterly phase = ACCUMULATION — improves win rate from 32.0% — E8 Intelligence Research
Abstract
DECISION: REJECT ELEMENT: No element survives — the ACCUMULATION filter is a regime label, not a timing edge, and cannot be grafted onto E8's geometry without redefining the node's phase logic. EXPECTED IMPACT: Win-rate gain is real but misleading — +12.2pp on n=120 is within noise for quarterly splits (4–5 quarters per bucket). Even if true, the filter cuts trade frequency by ~50–80% (ACCUMULATION is rare), likely reducing total R despite higher win rate. The live-window "improvement" (+2R) is 4 trades — not a signal. RISK: Overfitting to a single phase label; ACCUMULATION quarters often precede high-volatility breakouts, so filtered-out trades may be the highest-R winners. Also, the filter ignores R-multiple distribution — a 44% win rate with 1.5R average loss vs 1.0R average win is still negative expectancy. TEST: Before any reconsideration, run a 24-month replay-bench on all rizzy5m, rizzy, sqz, wkjudas signals, splitting by quarter phase. Compute profit factor and average Author: Andrew Stewart Caldin, Independent Researcher, UK. Part of the E8 Intelligence Research series. Platform: e8intelligence.com
