A trading strategy can show a high win rate and still lose money, or produce strong returns while exposing an account to unacceptable drawdowns. This guide explains the main trading strategy performance metrics—including expectancy, profit factor, maximum drawdown, Sharpe ratio, Sortino ratio, exposure, turnover, and benchmark-relative return—and provides a repeatable checklist for evaluating backtests, trading bots, and live results.
Overview
No single statistic can prove that a strategy is reliable. Performance analysis works best when metrics are read together and connected to a clearly defined trading process. Before reviewing results, record the market, instrument, timeframe, entry and exit rules, position-sizing method, fees, slippage assumptions, leverage, and test period. Without this context, even accurate calculations can be misleading.
Separate three questions during the review:
- Is the strategy profitable? Examine net profit, expectancy, average trade, and profit factor after realistic costs.
- Is the path acceptable? Examine maximum drawdown, duration of drawdowns, volatility, and recovery factor.
- Does the return justify the risk and effort? Examine Sharpe, Sortino, exposure, turnover, benchmark-relative return, and operational complexity.
Use consistent conventions. For example, decide whether returns are measured per trade, per day, or per month; whether results are arithmetic or compounded; and whether commissions, borrow costs, funding, spreads, and market impact are included. A trading bot performance report should also distinguish realized results from open-position profit and loss.
Core trading strategy performance metrics
| Metric | Definition or formula | What it helps answer | Important limitation |
|---|---|---|---|
| Win rate | Winning trades divided by total trades | How often trades close profitably | It says nothing about the size of wins versus losses |
| Average win | Total winning profit divided by winning trades | How much a typical winner contributes | Large outliers can distort the average |
| Average loss | Total losing loss divided by losing trades, usually shown as a positive amount | How costly a typical losing trade is | It may hide a small number of extreme losses |
| Expectancy | (Win rate × average win) − (Loss rate × average loss) | Expected profit or loss per trade before considering uncertainty | It is an estimate, not a promise |
| Profit factor | Gross winning profit divided by gross losing loss | Whether total gains exceed total losses | It can look strong with too few trades or without costs |
| Maximum drawdown | Largest peak-to-trough decline in the equity curve | The worst historical loss from an account high | Future drawdowns can be larger |
| Recovery factor | Net profit divided by maximum drawdown | How much profit was produced relative to the worst decline | It depends heavily on the selected test window |
| Sharpe ratio | Average excess return divided by return volatility | Return relative to overall variability | It treats upside and downside volatility similarly |
| Sortino ratio | Average excess return divided by downside deviation | Return relative to harmful volatility | Results depend on the target return and downside convention |
| Exposure | Share of time, capital, or portfolio risk committed | How continuously the strategy is at risk | Different platforms calculate it differently |
| Turnover | Trading activity relative to capital or portfolio value | How frequently positions are changed | High turnover can make costs and execution decisive |
| Benchmark-relative return | Strategy return compared with a relevant passive benchmark | Whether active decisions added value | The benchmark must match the strategy's market and risk profile |
Expectancy is especially useful because it combines win rate and payoff size. A strategy with a 40% win rate can be profitable if its average winners are sufficiently larger than its average losses. Conversely, a strategy with frequent small wins may fail when one large loss overwhelms them. Calculate expectancy using net results where possible, and report the number of trades behind the estimate.
Checklist by scenario
For an initial backtest
- Define the test universe, timeframe, entry rules, exits, position sizing, and rebalancing schedule.
- Use data that was available at the time each decision would have been made. Check for survivorship bias, look-ahead bias, corporate-action errors, and missing prices.
- Include estimated commissions, spreads, slippage, borrow costs, financing, and other applicable trading expenses.
- Review the equity curve, monthly or weekly returns, trade distribution, and longest losing streak—not only the final return.
- Separate in-sample development data from out-of-sample validation data. A final untouched period can provide a further check against overfitting.
- Record the exact rules and software settings so the result can be reproduced.
For more on test design, see Backtesting Trading Strategies: A Practical Guide to Avoiding False Results.
For comparing two strategies
- Compare them over the same dates, instruments, data frequency, cost assumptions, and capital constraints.
- Start with drawdown and downside risk, then examine return metrics. A higher return may simply reflect higher leverage or greater exposure.
- Compare expectancy and profit factor with trade counts. A small sample should not be treated as decisive.
- Check whether returns come from one market regime, one instrument, or a handful of unusually large trades.
- Compare benchmark-relative performance and correlation. Two strategies with similar returns may provide very different diversification benefits.
When strategies rotate among sectors or ETFs, pair these metrics with trend and flow analysis. The ETF Rotation Strategy Guide provides a useful framework for that additional review.
For monitoring a live trading bot
- Compare live fills with the backtest's assumed fills, including spread and slippage.
- Track realized and unrealized profit separately, along with open risk, exposure, leverage, and cash usage.
- Monitor daily and cumulative drawdown, error rates, missed orders, rejected orders, and connectivity events.
- Measure performance by strategy version so a rule change does not obscure deterioration.
- Set pre-defined review thresholds, but avoid changing rules solely because of a short run of losses.
A practical monitoring workflow is outlined in the pre-deployment and live-monitoring checklist for trading bots.
For a strategy affected by market catalysts
Tag trades by conditions such as earnings, economic releases, high-volatility sessions, or quiet periods. Then compare expectancy, drawdown, exposure, and execution quality for each group. This can reveal that a strategy is profitable only during a narrow type of market. The economic data trading guide and earnings-gap checklist can help define those event categories without mixing fundamentally different setups.
What to double-check
Risk-free rate and return frequency: Sharpe and Sortino calculations require a return benchmark or target. State the risk-free-rate assumption, the return interval, and the annualization method. Annualizing daily data is not automatically appropriate if returns are autocorrelated or the strategy has irregular exposure.
Drawdown definition: Confirm whether maximum drawdown is based on closing equity, intraday equity, realized results, or total equity including open positions. Also record drawdown duration and time to recovery. Two strategies can have the same maximum drawdown but very different experiences.
Costs and execution: A backtest that uses idealized fills may overstate results, particularly for high-turnover strategies, thinly traded securities, crypto markets, or news-driven moves. Recheck spread assumptions when the strategy trades outside normal liquidity conditions.
Benchmark selection: Compare a broad stock strategy with a relevant broad benchmark, an ETF strategy with an appropriate ETF benchmark, and a market-neutral strategy with a cash or low-risk reference where appropriate. Explain differences in leverage, exposure, dividends, and volatility.
Distribution of outcomes: Inspect median trade, largest winner, largest loser, losing streak, monthly dispersion, and tail results. Mean values can hide a strategy that depends on a single exceptional trade.
Regime sensitivity: Group results by rising, falling, sideways, high-volatility, and low-volatility periods where the data allows. A strategy need not perform equally in every environment, but its weak conditions should be known before capital is allocated. See How to Identify Market Regimes and Choose the Right Trading Strategy for a structured approach.
Common mistakes
- Using win rate as the headline metric: Always pair it with average win, average loss, expectancy, and the largest loss.
- Ignoring the sample size: A short track record can produce unstable ratios and an unusually favorable equity curve.
- Optimizing until the backtest looks perfect: Excessive parameter searches increase the chance that rules fit historical noise rather than a repeatable effect.
- Reporting only percentage returns: Include starting capital, leverage, exposure, drawdown, costs, and the benchmark.
- Mixing realized and open results: Mark-to-market losses and gains affect actual risk even before a position is closed.
- Changing the strategy during a drawdown without evidence: First check implementation, data, execution, and whether the drawdown is within the tested distribution.
- Comparing incompatible reports: Different periods, fees, currencies, compounding methods, or benchmark definitions can make a numerical comparison meaningless.
Position sizing often determines whether a statistically attractive strategy is usable in practice. Review the position-sizing guide before translating a backtest into live capital.
When to revisit
Review the full metric set at the end of each month or quarter, after a meaningful strategy revision, and whenever a drawdown approaches a pre-defined risk limit. Recalculate results after changing the broker, data feed, execution venue, fees, position-sizing rules, or trading software. For seasonal planning, refresh the data, benchmark, risk-free-rate assumption, and market-regime labels before deciding whether to increase, reduce, or pause exposure.
Use this action sequence:
- Export the latest trades and equity curve.
- Recalculate net expectancy, profit factor, drawdown, recovery factor, Sharpe, Sortino, exposure, turnover, and benchmark-relative return using documented conventions.
- Compare live results with the original backtest and the most recent review period.
- Investigate differences by instrument, setup, market regime, execution quality, and strategy version.
- Decide in advance whether the evidence supports continuing, reducing risk, pausing for investigation, or retiring the strategy.
- Record the decision and the next review date in a trading journal.
Metrics are decision tools, not guarantees. The strongest review combines statistical results with realistic execution assumptions, transparent data, and a risk level that fits the trader's tolerance. Revisit the checklist whenever the underlying inputs change rather than waiting for a large loss to force the review.