Is your backtest telling the truth?

The more parameter combinations you try, the better the best one looks — even on pure noise. This is the single most common reason a strategy that "worked" in testing loses money live. The correction below deflates a Sharpe ratio by how hard you searched for it.

The checklist that actually matters

• Walk-forward, not one backtest. Optimise on old data, test on the next unseen slice — repeatedly.
• Out-of-regime. A strategy tuned in a bear year must be tested in a bull and a chop. We ran ours in four regimes; the design-window winner lost in all three others.
• Profit factor and win/loss ratio, never win rate alone. 85% win rates lost us money; a few big losses erased many small wins.
• Stress the costs. Re-run at 2x and 3x fees. If the edge dies, it was never an edge.
• Write the kill criterion before you look at the result. Everything else is negotiation with yourself.