Testing tells you how a strategy would have behaved. It is essential — and easy to fool yourself with. This page covers how to test and, more importantly, how to read the results without lying to yourself.
Backtesting#
A backtest runs your strategy against historical market data and reports how it would have performed. It answers "did these rules make sense in the past?" A good backtest models the frictions that real trading has:
- Fees on every entry and exit.
- Slippage — the gap between the price you expected and the price you got.
- Realistic fills — not assuming you always got the perfect price.
A backtest that ignores fees and slippage will look far better than reality.
Reading results honestly#
Look past the headline return. The numbers that tell you whether a strategy is survivable:
- Maximum drawdown — the deepest peak-to-trough loss. Could you stomach it live?
- Number of trades — ten trades prove nothing; a few hundred start to mean something.
- Win rate and reward-to-risk together — a 40% win rate can be very profitable with the right payoff, and a 70% win rate can lose money with the wrong one.
- Consistency — is the result spread across the whole period, or driven by one lucky stretch?
The curve-fitting trap#
Curve-fitting is tuning a strategy until it looks perfect on past data — and useless on future data. Warning signs:
- Lots of parameters, each tuned to a suspiciously specific value.
- Results that fall apart if you change a setting slightly.
- A strategy that only ever worked in one market during one period.
Guard against it: keep rules simple, prefer strategies that stay robust when you nudge the inputs, and validate on data the strategy wasn't tuned on. A slightly worse backtest that holds up out-of-sample beats a perfect one that doesn't.
Then: paper trade#
A backtest is history. Before real capital, run the strategy forward in paper mode on live data — see Paper Trading and Deploying Live vs Paper.