Strategy Validation Plan
0) Pre-commit (avoid hindsight bias)
- Pre-register exact rules (entries/exits/sizing) and tag the repo (e.g.,
v0.9-pretrial
). - Freeze data provider/version; document fees, slippage, funding (for perps).
1) Empirical Validation (backtest/out-of-sample)
- Time-series split: in-sample 60–70%, validation 15–20%, out-of-sample 15–20%.
- Walk-forward test; stitch OOS segments.
- Metrics: Sharpe ≥ 1.0, Sortino ≥ 1.2, PF ≥ 1.2, MaxDD ≤ 25%, Calmar ≥ 0.4.
- Overfitting checks: sensitivity heatmaps, Deflated Sharpe, SPA/Reality Check, Monte-Carlo equity cones.
- Regime robustness: bull, chop, crash slices.
2) Paper Trading (live sim)
- Run 4–8 weeks 24/7 with real quotes; log slippage vs assumptions, rejects, outages.
- Pass if Sharpe ≥ 0.7, PF ≥ 1.1, MaxDD within +10% of backtest OOS; ops error < 2% orders.
3) Pilot With Tiny Capital
- Start with 0.5–1.0% of intended capital; no leverage initially.
- Risk with ATR or vol targeting; daily loss cap; kill-switch at DD thresholds.
- Pass if real PF ≥ 1.15, Sharpe ≥ 0.7, tracking error vs paper/backtest ≤ 25%.
Deliverables (for class & repo)
- Methodology brief (rules, data, slippage/fees).
- Validation report (equity/DD, rolling Sharpe, sensitivity, OOS, regime slices, overfit tests).
- Operations appendix (latency, order types, failure handling, audit trail).
- Go-live plan (paper criteria, pilot size, kill-switch, monitoring).
- One-page summary with acceptance criteria & results.