Backtesting Guide
Backtests are evidence, not proof. Use them to test rules and avoid story-driven decisions.
Core idea: A good backtest is repeatable, interpretable, and robust to small parameter changes.
What to measure
- ROI / CAGR (return) and max drawdown (risk).
- Win rate and avg win vs avg loss (edge shape).
- Exposure and turnover (realism/fees sensitivity).
Common pitfalls
- Overfitting: too many rules tuned to the past.
- Lookahead bias: using information not available at the time.
- Survivorship bias: ignoring delisted/failed assets.
Validation steps (practical)
- Test multiple markets/regimes (bull/bear/high vol).
- Use walk-forward or holdout periods.
- Stress test assumptions (slippage/fees) and avoid fragile edges.