← Back to /audit
Anti-Overfit Audit — Deflated Sharpe Ratio per Strategy
RESEARCH SURFACE — ADVISORY ONLY. NOT A PRODUCTION GATE. NOT FINANCIAL ADVICE.
Live data: audit_dashboard/data/anti_overfit_audit.json · Source: ejaguiar1_stocks::trading_picks closed picks · Auto-refresh: hourly via .github/workflows/audit-dashboard.yml · Validator: alpha_engine.anti_overfit_validator.deflated_sharpe (Lopez de Prado AFML eq 14.5) · SUPREME EDGE P1 wire-up (2026-05-11)
Thresholds
| Verdict | DSR range | Interpretation |
|---|
| EDGE_LIKELY_REAL | ≥ 0.95 | Headline Sharpe survives multiple-testing correction (Bailey-Lopez de Prado "publishable confidence"). Promotion candidate to OOS_READY in state machine if other gates pass. |
| UNDETERMINED | 0.50 — 0.95 | Statistically inconclusive given sample + skew/kurtosis. More data needed. |
| OVERFIT_LIKELY | < 0.50 | Backtest-overfit signature; reject or quarantine. Most likely the strategy's positive Sharpe is selection bias, not real edge. |
Per-strategy verdicts
Loading...