A backtest applies fixed rules for entry, exit and sizing to past data, trade by trade, and reports results such as win rate, expectancy, profit factor and max drawdown. It can be automated in code or done by hand, replaying charts candle by candle. A useful backtest includes realistic fees, slippage and funding.
Example
You test a rule: buy BTC when a four-hour candle closes above the 20-period high, with a stop at 1.5× ATR and a 3R target. Over two years of data it produces 140 trades, a 33% win rate and an expectancy of +0.25R after fees, with a max drawdown of 14R. That justifies trading it small and comparing the live results with the test.
Why it matters
Backtesting shows whether an idea has any edge before you pay to find out live. It also sets expectations: the losing streaks and drawdowns in the test help you stick with the strategy when the same thing happens in real trading.
Common mistake
Tweaking parameters until the backtest looks perfect. Optimising on the same data you test on, known as overfitting, produces results that rarely survive live trading. Keep the rules simple and check them on data you did not use to build them.
Put it into practice
FAQ
- How many trades does a backtest need?
- At least 100 trades is a common minimum for meaningful statistics. More is better, ideally across different market conditions.
- What is overfitting?
- Fitting a strategy so closely to past data that it captures noise instead of a real pattern. It shows up as great backtests followed by poor live results.
- Is manual backtesting worth it?
- Yes, especially for discretionary setups that are hard to code. Replaying charts candle by candle also trains your pattern recognition.