How to Read Backtest Results Like a Professional

A backtest report is a dense page of numbers, and the numbers are not equally important. Reading one well means knowing what each metric measures, what range is normal, and, above all, which combinations contradict each other. This lesson is the field guide: the previous backtesting lesson covered how to run an honest test; this one covers how to read what it prints.

The return numbers

Total return is what the strategy made over the whole test. On its own it is nearly meaningless: 60 percent is superb over two years and mediocre over twelve.

CAGR, the compound annual growth rate, fixes that by converting any test length to an equivalent yearly rate, which is the number to compare against alternatives. The honest benchmark is not zero; it is what simply buying and holding the same instrument returned over the same window. A strategy that made 12 percent a year trading an asset that made 15 percent by sitting still earned a negative wage for its complexity.

The pain numbers

Maximum drawdown is the deepest peak-to-trough fall in the equity curve, and it is the single most predictive number for whether you will actually stick with the strategy. You will live through a drawdown like it again, in real time, without knowing it will end.

Two habits make drawdown useful. First, read it against CAGR: a strategy compounding at 20 percent with a 35 percent max drawdown recovers its worst fall in about two years of average performance, which is a long time to keep faith. Second, read the drawdown's duration, not just its depth. A 20 percent hole that lasted a month reads very differently from the same depth stretched across a year and a half.

Volatility measures how bumpy the ride is day to day. It matters mostly through the ratios below and through position sizing: a more volatile strategy must trade smaller to hold the same account-level risk.

The efficiency ratios

Sharpe ratio is return per unit of volatility. Rough compass for retail backtests: below 1.0 weak, 1.0 to 2.0 respectable, above 2.0 strong, and far above that a reason for suspicion rather than celebration, because multi-year retail backtests rarely earn such numbers honestly.

Sortino ratio is the same idea counting only downside volatility, which forgives strategies whose bumpiness is mostly upward. It typically reads higher than Sharpe; compare each against its own scale, not against the other.

Profit factor is gross profit divided by gross loss. Below 1.0 the strategy lost money; 1.2 to 1.5 is workable if trade count is high; above 2.0 is excellent and, past a point, another number to double-check rather than admire.

The character numbers

Win rate describes the strategy's personality, not its quality. Trend followers commonly win 30 to 40 percent of trades and profit because winners dwarf losers; mean-reversion systems commonly win 60 to 80 percent with the opposite shape. The number that reconciles win rate with outcome is the ratio of average win to average loss, and the two together determine expectancy, the average profit per trade.

Concretely: a 40 percent win rate with winners 3 times the size of losers earns 0.6 units per trade risked. A 75 percent win rate whose rare losses are 5 times the average win loses money. Neither win rate alone told you that.

Trade count is the sample size behind every other number. Under about 30 trades the report is an anecdote; a few hundred trades give the statistics room to mean something.

Five cross-checks before trusting a report

1. Equity curve shape. Steady climb with breathing room is what real edges look like. One vertical jump carrying the whole return means the result is one event, not a system.
2. Returns by year. A strategy that made everything in 2021 and nothing since did not pass a five-year test; it passed a one-year test with four years of padding.
3. Costs included? Confirm fees, spread, and slippage were modeled. High trade counts with hairline profit factors flip sign on costs alone.
4. Parameter neighbors. If the reported settings work and both neighbors lose, the result is a coordinate, not an edge; robust strategies degrade gently as parameters shift.
5. Out-of-sample behavior. The clinching evidence: performance on data the development process never touched, whether a held-out recent window or a proper walk-forward test.

Every Horizon backtest lays these ingredients out for inspection: the equity curve with its drawdown profile, the metrics panel, and the trade list underneath it all. The strategy templates are a good place to calibrate your eye, since each publishes its rules next to its historical results, and the honest ones do not all look impressive.

Frequently asked questions

What is a good Sharpe ratio for a trading strategy?

For a retail strategy over a multi-year backtest, above 1.0 is respectable and above 2.0 is strong. Sustained Sharpe ratios far above 2.0 are rare enough in honest retail tests that the first response should be checking for overfitting or missing costs.

What is a good win rate for a trading strategy?

There is no good win rate in isolation. Profitable trend-following systems often win under 40 percent of trades, and losing systems can win over 80 percent. Judge win rate together with the average win to average loss ratio; the pair determines whether the strategy makes money.

What is a good maximum drawdown?

Smaller is better, but the practical test is proportional: a drawdown you would actually sit through, and one the return justifies. Many practitioners want max drawdown no worse than the CAGR, and treat anything beyond about a third of the account as behaviorally unrealistic whatever the math says.

How many trades does a backtest need to be reliable?

Treat 30 as the floor below which results are noise, and a few hundred as where statistics become comfortable. More trades also mean costs were tested realistically, since cost errors compound with frequency.