Hit rate, also called win rate, is the fraction of eligible outcomes that count as successes. In a trading report, it commonly means the fraction of completed trades with positive profit and loss, measured on a stated cost basis. It tells you how often a trade won; the sizes of the wins and losses decide whether those trades made money.
First decide what counts as a trade
For a closed-trade hit rate, let be the number of eligible completed trades and the number whose net P&L is strictly positive. Then
Multiplying by expresses the fraction as a percentage. With no eligible trades, the statistic is undefined, rather than zero. This is the same successes-over-trials calculation used for a sample proportion.
A reproducible report needs a counting contract:
- Unit: here, one trade is a complete round trip from no position, through an exposure, back to no position. Partial fills and scale-outs belong to that round trip. A multi-leg strategy needs an explicit rule for grouping its legs; counting each fill or leg separately answers a different question. Trading engines can offer different trade-grouping methods.
- Eligibility: specify the instruments, strategy, reporting window and time zone. Here, include round trips whose final exit falls inside the window, including any eligible trade opened earlier.
- Cost basis: use P&L after the specified transaction costs and holding charges. A gross winner can become a net loser. Do not deduct costs again from an already-net ledger.
- Zero and open outcomes: here, an exactly break-even closed trade stays in but not in . An open position is excluded until its round trip closes. If instead zero outcomes are excluded, label the result as wins divided by wins plus losses and disclose the zero count.
A closed-trade statistic can look flattering when losing positions are kept open. Report open positions and their marked P&L alongside it; the closed-trade count is not a complete portfolio valuation. Likewise, a positive-day rate or the fraction of correct direction forecasts uses a different unit and success rule. Neither is automatically the closed-trade win rate.
Three winners, yet a loss
Consider these five hypothetical completed round trips. P&L is in USD. Commission and execution-price effects are already included, with no financing charges, borrow charges, distributions or taxes in this example.
| Trade | Net P&L (USD) | Counts as a win? |
|---|---|---|
| A | +10 | Yes |
| B | −30 | No |
| C | +5 | Yes |
| D | +8 | Yes |
| E | −2 | No |
The hit rate is . Winners contribute USD, while losses total USD in magnitude, leaving −9 USD overall. No starting capital is specified, so this is a cash P&L example, not a portfolio return percentage.
Original synthetic closed-trade ledger in USD, already net of the specified costs. Each bar is one completed round trip; counting winners gives every trade one vote, regardless of its P&L.
Average winning P&L is USD; average loss magnitude is USD. The mean P&L across the five trades reconciles exactly:
This is a historical sample identity, not a forecast of expected return. Hit rate discards payoff size: it gives the small gain on C the same vote as the larger gain on A, and treats both losses simply as failures. Pair it with average win, average loss, total net P&L and the risk and capital required to produce them. A majority of winners is not a profitability test.
A frequency is not the next trade's probability
The five-trade ledger establishes only what happened in those five completed trades. It does not establish a probability of winning the next trade. A binomial model can describe independent success/failure trials with a common win probability; NIST's binomial reference makes that fixed-probability assumption explicit. Overlapping trades, shared market shocks and changing strategy conditions can defeat those assumptions. More trade records do not automatically mean equally many independent observations.
Keep the trade count next to the percentage, examine genuinely held-out backtesting results, and avoid selecting a strategy solely because its historical win rate looks best. That selection can introduce overfitting.
If a strategy also outputs win-probability forecasts, the next task is to compare those forecasts with matured outcomes on held-out data. The Probability calibration lesson covers that check before forecasts enter sizing; full lesson access requires membership, with a public preview. Calibration of forecasts does not by itself establish profitability.
Sources
- QuantConnect, Trade Statistics: completed-trade records and configurable grouping of fills. Its win-rate definition uses profitable trades after transaction fees.
- NIST Dataplot, Binomial Proportion: successes divided by trials.
- NIST/SEMATECH, Binomial Distribution: fixed success probability and the count model, distinct from an observed proportion.
- Ernest P. Chan, Quantitative Trading: How to Build Your Own Algorithmic Trading Business, Wiley, first edition, copyright 2009, Chapter 2, pp. 22–23: the effect of transaction costs on strategy performance. The example here is original.