A backtest asks a simple question: how would a defined trading strategy have behaved on historical data? A useful backtest is more than a profitable equity curve. It should make the rules explicit, include realistic costs and separate development data from data used for final evaluation.
Write the rules first
Before testing, define the entry, exit, stop, target, position-sizing rule, trading hours and conditions that prevent a trade. If the rules change after looking at the results, the test becomes vulnerable to overfitting.
Include costs and execution assumptions
Spread, commission and slippage can materially change short-term strategies. Tests should use assumptions that resemble the intended broker and instrument. A strategy that only works before costs deserves additional scrutiny.
Separate development and out-of-sample data
One common approach is to use an earlier period for model development and a later unseen period for evaluation. Walk-forward testing can repeat this process across multiple periods. The purpose is to test whether the rules remain useful when conditions change.
Look beyond win rate
Win rate alone does not describe a trading system. Average win, average loss, payoff ratio, drawdown, trade frequency, exposure and the distribution of outcomes all matter. A strategy with a high win rate can still lose money if occasional losses are much larger than winners.
Watch for overfitting
Adding parameters until historical performance looks excellent can produce a model that fits noise. Simpler rules, economic rationale, stability across instruments and out-of-sample testing provide stronger evidence than a highly optimized historical curve.
Key takeaway
A credible backtest is a research process, not a sales statistic. The goal is to understand how a strategy behaves under different conditions and how uncertain the historical result may be.
Related GodzillaBTC resources: Trading calculators · Signals · Methodology · Risk disclosure