How to Read a Backtest Report Like a Pro
You ran a backtest and got six numbers. They look fine. But which one matters most depends entirely on what you are trying to build — and most beginners look at the wrong one first.
Total Return Is Not the Goal
The first number anyone looks at is total return. It is the least useful number on the page. Two strategies can both return 20% over the same period and represent completely different risk profiles. Total return tells you what happened. It does not tell you whether you would have stayed in the trade.
Sharpe Ratio: Risk-Adjusted Reality
The Sharpe ratio divides excess return by the volatility of those returns. It answers: "how smooth was the ride."
A Sharpe under 1.0 in a backtest is unremarkable — most simple strategies clear this bar. A Sharpe above 1.5 is interesting. A Sharpe above 2.5 in a backtest is suspicious and almost always means look-ahead bias or curve-fit parameters. Real-world out-of-sample Sharpes for retail strategies tend to live between 0.7 and 1.5.
Max Drawdown: The Number That Tests Your Conviction
Max drawdown is the worst peak-to-trough loss the strategy ever went through. It is also the number that decides whether you actually run the strategy.
A 25% drawdown sounds tolerable on a spreadsheet. Sitting through it with real money — watching $50,000 turn into $37,500 over six weeks — feels like a different experience entirely. If you would not have stayed in, the strategy's reported return is irrelevant.
Win Rate vs. Win Size
A 70% win rate sounds excellent until you discover the average winner is half the size of the average loser. Win rate alone is meaningless without average win and average loss alongside it.
Trend strategies often have win rates of 40% with average winners three times larger than losers. Mean reversion strategies might have 70% win rates with smaller winners. Either profile can work — neither is automatically better.
Trade Frequency and Edge
A strategy that trades twice and produces 40% return is statistically meaningless. A strategy that trades 200 times and produces 12% return tells you something — there are enough samples to believe the result.
More trades makes the statistical signal stronger, but also multiplies costs. The right number depends on the timeframe and asset class. As a rough guide, fewer than 20 trades over a year of backtesting is usually too few to draw conclusions.
Putting It Together
Read the metrics in this order: drawdown first (could you sit through it), Sharpe second (was the ride reasonable), trade count third (is there enough signal), then win rate and average trade. Total return is a side effect. If those four numbers look good, the return number will take care of itself.