A senior-level guide to reading backtest metrics in the right order: validate the test first, screen profitability, inspect drawdown, then compare risk-adjusted returns.
從無程式的加密貨幣現貨策略開始,鎖定版本、執行回測,並讓結果保持可追溯以供比較。
The first backtest metric to trust is not Win Rate. If you are choosing between Win Rate, Profit Factor, and Sharpe Ratio, start with Profit Factor because it shows whether gross profits were large enough to cover gross losses.
But a professional review starts one step earlier: is the backtest reliable enough for any metric to matter?
Before ranking metrics, check trade count, fees, slippage, date range, market regime, execution assumptions, and whether the result holds beyond the exact period used to tune the strategy. A clean Win Rate, Profit Factor, or Sharpe Ratio can still be meaningless if the test is under-sampled, overfit, or based on unrealistic costs.
Backtesting is research evidence, not proof of future performance.
Use this order when reviewing a strategy backtest:
Among Win Rate, Profit Factor, and Sharpe Ratio, Profit Factor usually matters first for strategy screening. Win Rate is the easiest to misunderstand. Sharpe Ratio is most useful after the strategy has already passed basic profitability and risk checks.
Win Rate measures the percentage of trades that closed profitably.
Win Rate = Winning Trades / Total Trades
A 70% Win Rate looks strong, but it says nothing about the size of the wins or the size of the losses.
Example:
| Outcome |
|---|
This strategy wins 70% of the time and still loses money overall.
The reverse can also be true. A trend-following strategy may win only 35% or 40% of the time, but if average winners are much larger than average losers, the result can still be positive.
Use Win Rate to understand trade behavior, not strategy quality by itself.
Win Rate helps answer:
Win Rate does not answer:
Profit Factor compares total gross profit with total gross loss.
Profit Factor = Gross Profit / Gross Loss
A Profit Factor above 1.0 means the strategy made more from winners than it lost from losers during the test. A Profit Factor below 1.0 means losses exceeded gains.
As a rough interpretation:
Profit Factor matters because it connects Win Rate with payoff size. Two strategies can have the same Win Rate but completely different economics.
Same Win Rate. Very different strategy.
This is why experienced reviewers rarely evaluate Win Rate alone. Profit Factor gives a clearer view of whether the system turned risk into return efficiently.
Still, Profit Factor is not perfect. It can look excellent with too few trades, one unusually large winner, or an overfit parameter set. Always check trade count and the distribution of wins and losses.
Sharpe Ratio measures return relative to volatility.
Sharpe Ratio = (Portfolio Return - Risk-Free Rate) / Standard Deviation of Returns
A higher Sharpe Ratio generally means the strategy produced more return per unit of volatility.
Common rough interpretation:
| Sharpe Ratio | How to read it |
|---|
Sharpe is useful when comparing strategies with different volatility profiles. A lower-return strategy with a smoother equity curve may be more attractive than a higher-return strategy with violent swings.
But Sharpe has limits.
It can penalize upside volatility, depend heavily on the return sampling method, and hide tail risk. A strategy can show a good Sharpe Ratio and still suffer a sharp drawdown if losses cluster.
For trading systems with uneven payoff distributions, pair Sharpe with drawdown, Profit Factor, trade distribution, and outlier checks.
If the question is limited to the three metrics in the title, the ranking is:
Profit Factor comes first because it asks whether total wins outweighed total losses. Sharpe Ratio comes next because it shows whether the return path was efficient relative to volatility. Win Rate comes last because it can feel persuasive while leaving the economics incomplete.
The fuller professional answer is:
First validate the backtest. Then check Profit Factor and expectancy. Then inspect drawdown. Then use Sharpe Ratio for risk-adjusted comparison. Treat Win Rate as supporting context.
That order keeps you from falling in love with a headline number before understanding what created it.
Expectancy estimates the average expected profit or loss per trade.
Expectancy = (Win Rate x Average Win) - (Loss Rate x Average Loss)
Expectancy is the bridge between Win Rate and Profit Factor.
A strategy with positive expectancy can survive a lower Win Rate if winners are large enough. A strategy with negative expectancy can fail even with a high Win Rate if losses are too large.
Before trusting any backtest, ask:
Do not start with the best-looking metric. Start with the test.
Ask:
A backtest with 20 trades and a Profit Factor of 3.0 is less convincing than a backtest with hundreds of trades, realistic costs, and stable results across adjacent periods.
Next, review:
This tells you whether the strategy has a basic economic edge in the tested sample.
Then review:
A profitable strategy can still be unusable if the drawdown is too deep, too long, or too dependent on one recovery event.
Now Sharpe Ratio becomes useful.
Compare Sharpe across strategy variants only when the tests use the same market, timeframe, date range, and cost assumptions. Otherwise, the comparison can be misleading.
Finally, use Win Rate to understand how the strategy behaves in practice.
A low Win Rate strategy may be profitable but psychologically difficult. A high Win Rate strategy may feel comfortable but carry hidden crash risk. This matters because a strategy that cannot be followed consistently is not very useful.
The biggest mistake is treating one metric as the answer.
Other common mistakes include:
A good backtest should survive questions. If one question destroys the result, the strategy was weaker than the headline metric suggested.
When comparing strategy variants, do not ask which one has the highest ending balance first.
Use this checklist:
The best backtest is not always the one with the highest return. It is the one with the most credible tradeoff between profitability, risk, repeatability, and operational fit.
Traseq is a no-code crypto spot strategy research workspace. It is designed for building, backtesting, comparing, and tracking strategy versions before any live decision.
A practical Traseq workflow is:
Traseq is research software, not live trading or exchange execution. Backtest results help evaluate historical behavior; they do not guarantee future returns.
For workflow details, read the first backtest guide, Reading Backtest Results, and Comparing Multiple Backtests.
See the Traseq backtest workflow to build one version-linked strategy test, review the metrics, and compare the result against a cleaner alternative.
There is no single metric that explains everything. For first-pass strategy screening, Profit Factor and expectancy are usually more useful than Win Rate because they include the size of wins and losses. For risk review, maximum drawdown is essential. For risk-adjusted comparison, Sharpe Ratio is useful.
No. Win Rate only shows how often trades are profitable. Profit Factor shows whether total gains exceeded total losses. A high Win Rate strategy can still lose money if average losses are much larger than average wins.
A Profit Factor above 1.0 means gross profit exceeded gross loss in the tested period. Many traders treat 1.5 or higher as a stronger candidate, while values above 2.0 deserve attention. Extremely high Profit Factor can also be a warning sign if the sample is small or overfit.
A Sharpe Ratio above 1.0 is often considered solid, and above 2.0 is strong. But Sharpe should not be used alone because it can hide drawdown, tail risk, and unstable return patterns.
Yes. A low Win Rate strategy can be profitable if average winners are much larger than average losers. Many trend-following strategies work this way.
Yes. A strategy can win frequently but lose money if occasional losses are much larger than regular wins. This is why Win Rate must be reviewed with average win, average loss, Profit Factor, and expectancy.
There is no universal number, but very small samples are unreliable. As a practical rule, treat results with fewer than 100 trades carefully, and prefer larger samples when the strategy trades frequently enough to support them. Data quality, market regime coverage, and robustness testing still matter.
Do not optimize blindly for either. Profit Factor is useful for profitability screening, while Sharpe Ratio is useful for risk-adjusted comparison. A robust strategy should also pass drawdown, cost, sample size, and out-of-sample checks.
| Count |
|---|
| Average result |
|---|
| Total |
|---|
| Winners | 70 | $10 | $700 |
| Losers | 30 | -$50 | -$1,500 |
| Profit Factor | How to read it |
|---|
| Below 1.0 | Losing in the tested period |
| 1.0 to 1.3 | Weak or cost-sensitive |
| 1.3 to 1.5 | Potential candidate, needs validation |
| 1.5 to 2.0 | Stronger candidate |
| Above 2.0 | Strong, but check robustness |
| Above 4.0 | Be skeptical unless sample quality is excellent |
| Strategy | Win Rate | Average win | Average loss | Read |
|---|
| A | 60% | $50 | -$50 | Profitable profile |
| B | 60% | $50 | -$150 | Losing profile |
| Below 1.0 | Weak risk-adjusted return |
| 1.0 to 2.0 | Solid |
| 2.0 to 3.0 | Strong |
| Above 3.0 | Excellent, but check for smoothing or overfitting |
| Question | Why it matters |
|---|
| Same market, timeframe, and date range? | Keeps comparison fair |
| Same fee and slippage assumptions? | Prevents cost distortion |
| Better Profit Factor? | Screens profitability efficiency |
| Lower and shorter drawdown? | Screens survivability |
| Stable across periods? | Reduces regime dependence |
| Less dependent on outliers? | Reduces fragility |
| Better Sharpe Ratio? | Compares risk-adjusted return |
| Manageable Win Rate and losing streaks? | Checks usability |