A trading strategy that sounds logical and a trading strategy that's actually profitable are not the same thing, and the gap between them is exactly what backtesting is meant to reveal. Backtesting means running a strategy's rules against historical price data to see how it would have performed, before risking real capital on it. This entry breaks down the core metrics a backtest should produce and how to interpret them, since a backtest that only reports "it made money" leaves out most of the information that actually matters.
The Core Metrics Table
| Metric | What It Measures | Why It Matters |
|---|
| Win rate | Percentage of trades that were profitable | A low win rate isn't automatically bad if average wins are large enough relative to average losses — this metric means little in isolation |
| Average win / average loss ratio | How large winning trades are relative to losing trades | Combined with win rate, this determines whether the strategy is mathematically profitable over a large sample |
| Maximum drawdown | The largest peak-to-trough decline in account value during the test period | Reveals the worst-case emotional and financial stress the strategy would have required living through |
| Sample size (number of trades) | How many total trades the backtest is based on | A strategy tested on a small number of trades can look excellent purely by chance; results need enough trades to be statistically meaningful |
| Performance across different market regimes | How the strategy performed in trending versus range-bound versus high-volatility periods | A strategy that only works in one type of market condition carries hidden risk that a single aggregate result can hide |
Interpreting the Combination, Not Any Single Metric Alone
A high win rate paired with a poor average-win-to-average-loss ratio can still be a losing strategy overall, if the rare losses are large enough to outweigh the frequent small wins. Conversely, a lower win rate paired with a strong average-win-to-average-loss ratio can be solidly profitable. Neither win rate nor the win/loss ratio tells the full story in isolation — they need to be read together, alongside drawdown and sample size, before drawing any conclusion about whether a strategy is genuinely sound.
A Common Backtesting Trap Worth Naming Directly
A backtest that looks excellent is sometimes the result of curve-fitting — adjusting a strategy's specific rules and parameters repeatedly until they happen to fit the historical data extremely well, rather than because the underlying logic is genuinely sound. A strategy that's been tuned this way often performs noticeably worse on new, unseen data than the original backtest suggested, precisely because its apparent edge was fitted to noise in the historical sample rather than reflecting a real, repeatable pattern.
Turning the Read Into a Rule
A practical rule that follows from this: before trusting any backtest result, check that it covers a large enough sample size across multiple different market regimes, and be specifically skeptical of a strategy with an unusually high win rate and low drawdown achieved through heavy parameter tuning on a single historical period. Testing a strategy on a separate, later period of data than the one used to originally develop it — sometimes called out-of-sample testing — is one of the more reliable ways to check whether a backtest result reflects a genuine edge rather than overfitting.
Why This Framework Is Evergreen
The specific strategy being tested changes constantly, but the process for evaluating whether a backtest result is trustworthy doesn't. Win rate, win/loss ratio, drawdown, sample size, and regime consistency are the same five checks worth running regardless of what specific rules the strategy under test actually uses.
The Takeaway
A backtest that only reports a single aggregate profit number is incomplete. Reading win rate alongside the win/loss ratio, checking maximum drawdown, confirming an adequate sample size, and testing across multiple market regimes — plus a healthy skepticism toward results that look too good relative to how heavily the strategy's rules were tuned — separates a strategy with a genuine, testable edge from one that simply got lucky fitting historical noise.
This post is educational content for traders and not financial advice. Backtested performance does not guarantee future results, and even a well-tested strategy carries real risk when traded with actual capital. Trade with capital you can afford to lose.
Comments
Post a Comment