How to Calculate Win Rate in Trading, With Break-Even and Sample-Size Tables
The win rate formula and the counting rules that change the number. Tables computed on 22 October 2026 give break-even win rates from 0.5R to 3R after costs, 95% confidence intervals, the trades needed to confirm an edge and the losing-streak odds each win rate implies.
To calculate win rate in trading, divide the number of winning closed trades by all closed trades and multiply by 100: 22 winners out of 50 trades is a 44% win rate. The figure only means something once you compare it with the break-even rate for your reward:risk and check the sample is large enough to rule out luck.
Five steps take you from a trade log to a number you can use:
- Export every closed trade with its net profit or loss after costs (spread, commission, swap and any exchange fees).
- Mark each trade a win (net P&L above zero) or a loss (below zero), log trades that close at exactly zero as scratches, and write those rules down.
- Divide winning trades by all closed trades and multiply by 100.
- Work out expectancy from your average win and average loss, then compare your win rate with the break-even rate for your reward:risk.
- Put a 95% confidence interval around the win rate before you trust it.
AlphaScala computed the tables below on 22 October 2026. The method is stated under each one so you can reproduce them.
The win rate formula and the counting rules behind it
Win rate = winning closed trades ÷ total closed trades × 100
Loss rate is the same calculation with losing trades on top. If you record scratch trades separately, win rate and loss rate will not add up to 100%.
The arithmetic is trivial. What goes into the numerator and the denominator is where two traders using the same journal end up with different numbers from identical trades. TradingView's "Percent profitable" metric, for example, "only considers closed trades", so open positions are left out, and it keeps "even" trades in the denominator but out of the numerator (TradingView Help Center, "Percent profitable", accessed 22 October 2026). Whichever convention you pick, keep it fixed across every review period, or one month's 45% cannot be compared with the next month's.
Four counting rules settle most of the disagreements:
- Closed trades only. An open position has no final outcome, so including it makes the figure move with every tick.
- Net of costs. Classify a trade after every cost it paid, from spread and commission to swap and exchange fees. A trade that made $6 on price and paid $8 in commission is a loss.
- Scratches stay in the denominator. A trade that closes at exactly zero is a trade, not a win. With 21 wins, 27 losses and 2 scratches, win rate is 42.0% with scratches counted and 43.75% without them.
- One position, one trade. Merge scale-ins and partial exits that belong to the same plan into a single record, or a position closed in four pieces becomes four wins.
If your journal stores only entry and exit prices, the forex profit calculator converts entry, exit, lot size and direction into a cash P&L, which is the number the win/loss classification needs.
Expectancy and the break-even win rate for each reward:risk
Win rate says how often a strategy pays. Expectancy says how much it pays per trade, on average:
Expectancy = (win rate × average win) − (loss rate × average loss)
A positive figure means the trades in the sample made money on average; a negative one means they lost it. Expressing it in R makes results comparable across account sizes. R is the amount risked on one trade, which is the distance from entry to stop multiplied by position size. The position size calculator works out the lot size at which your stop costs exactly the amount you chose to risk, which keeps "average loss = 1R" true in the log.
Worked example: a 50-trade log
Take a hypothetical journal of 50 closed trades, each risking $100:
| Item | Value |
|---|---|
| Winning trades | 22 |
| Losing trades | 28 |
| Win rate | 22 ÷ 50 = 44% |
| Average win | $180 (1.8R) |
| Average loss | $100 (1R) |
| Expectancy | (0.44 × $180) − (0.56 × $100) = +$23.20 per trade, or +0.23R |
| Net P&L check | (22 × $180) − (28 × $100) = $1,160, which is $23.20 × 50 |
| Profit factor (gross profit ÷ gross loss) | $3,960 ÷ $2,800 = 1.41 |
This log lost more often than it won and still made money, because the average win is 1.8 times the average loss. The same arithmetic works against a high win rate. A strategy that wins 70% of the time with an average loss three times its average win has an expectancy of (0.7 × 1R) − (0.3 × 3R) = −0.2R per trade.
Planned reward:risk and realised reward:risk drift apart once stops slip and targets get cut short, so use the averages from closed trades here. The risk-reward ratio guide covers the planned version.
Break-even win rate from 0.5R to 3R
Setting expectancy to zero gives the win rate at which a strategy neither makes nor loses money. Write reward:risk as R, the average win divided by the average loss (1.8 in the worked example):
Break-even win rate = 1 ÷ (1 + R)
Costs shift that line. If every trade, winner or loser, pays a cost of c (measured in R), a win nets R − c and a loss costs 1 + c, so the break-even rate becomes:
Break-even win rate with costs = (1 + c) ÷ (1 + R)
| Reward:risk (average win ÷ average loss) | Break-even, no costs | Costs of 0.05R per trade | Costs of 0.10R per trade |
|---|---|---|---|
| 0.5R | 66.7% | 70.0% | 73.3% |
| 0.75R | 57.1% | 60.0% | 62.9% |
| 1R | 50.0% | 52.5% | 55.0% |
| 1.25R | 44.4% | 46.7% | 48.9% |
| 1.5R | 40.0% | 42.0% | 44.0% |
| 2R | 33.3% | 35.0% | 36.7% |
| 2.5R | 28.6% | 30.0% | 31.4% |
| 3R | 25.0% | 26.3% | 27.5% |
Method: the break-even rate p solves p(R − c) − (1 − p)(1 + c) = 0. Costs are a flat charge per trade in R units. That fits spread and commission better than slippage, which varies from trade to trade.
Cost in R depends on stop distance. A 1-pip round-trip cost on a trade with a 20-pip stop is 0.05R; the same cost on a 10-pip stop is 0.10R. At 1R, break-even is 52.5% with the 20-pip stop and 55.0% with the 10-pip stop, against 50.0% before costs. The pip value calculator shows what one pip is worth for a given pair and lot size, which turns a pip cost into cash.
How many trades before a win rate means anything
A win rate from 20 trades is a rough estimate of the strategy's true win rate, and a confidence interval shows how rough. The NIST/SEMATECH e-Handbook of Statistical Methods (section 7.2.4.1) gives the common normal-approximation interval and the Wilson score interval. In a 2001 Statistical Science paper, Lawrence Brown, Tony Cai and Anirban DasGupta said common textbook prescriptions about the standard interval's safety were "misleading and defective in several respects". They recommended "the Wilson interval or the equal-tailed Jeffreys prior interval for small n" and the Agresti-Coull interval for larger n. AlphaScala used the Wilson interval for both interval tables below; at an observed 50% win rate the Wilson and Agresti-Coull intervals give identical bounds.
| Closed trades | 95% Wilson interval for an observed 50% win rate | Margin (± points) | Chance a true 50% strategy shows 60% or better |
|---|---|---|---|
| 20 | 29.9% to 70.1% | ±20.1 | 25.2% |
| 50 | 36.6% to 63.4% | ±13.4 | 10.1% |
| 100 | 40.4% to 59.6% | ±9.6 | 2.8% |
| 300 | 44.4% to 55.6% | ±5.6 | 0.03% |
| 500 | 45.6% to 54.4% | ±4.4 | under 0.01% |
| 1,000 | 46.9% to 53.1% | ±3.1 | under 0.01% |
Method: Wilson score interval with z = 1.96 (two-sided 95%), using the formula in NIST/SEMATECH section 7.2.4.1; every interval was cross-checked against the Wilson option of SciPy 1.16.3's binomial test. The last column is the exact binomial probability of at least 0.6 × n wins when the true win rate is 50%.
After 20 trades, an observed 50% is consistent with a true win rate anywhere from about 30% to 70%. On a 1:1 strategy that range runs from a heavy loser to a comfortable winner.
Twenty trades cannot tell those apart.
The same calculation run the other way: a strategy whose true win rate is 60% shows 50% or worse over 20 trades 24.5% of the time, and a strategy with no edge at 50% shows 60% or better 25.2% of the time. After 100 trades the interval is still almost ten points wide on each side.
Trades needed for the win rate to clear break-even
A more useful test than "is the win rate high" is "is the lower end of the interval above break-even". This table gives the number of closed trades needed before the 95% Wilson lower bound clears the no-cost break-even rate, assuming the observed win rate holds steady as trades are added:
| Reward:risk | Break-even win rate | Observed win rate | Trades needed |
|---|---|---|---|
| 1R | 50.0% | 55% | 385 |
| 1R | 50.0% | 60% | 97 |
| 1.5R | 40.0% | 45% | 369 |
| 1.5R | 40.0% | 50% | 93 |
| 2R | 33.3% | 40% | 193 |
| 2R | 33.3% | 45% | 63 |
| 3R | 25.0% | 30% | 289 |
| 3R | 25.0% | 35% | 73 |
Method: smallest n for which the Wilson 95% lower bound, at the observed rate, exceeds 1 ÷ (1 + R). Wins are counted as win rate × n, so the counts are approximate; requiring a whole number of wins (rounded down) adds 1 to 9 trades per row. No trading costs; with costs, break-even rises (see the table above) and the counts go up.
The cushion over break-even matters more than the win rate itself. Running the same test for every reward:risk in the break-even table, from 0.5R to 3R, a win rate 5 points above break-even needs 289 to 385 trades to confirm, and one 10 points above needs 73 to 97.
Back to the worked example. Its 22 wins in 50 trades give a Wilson interval of 31.2% to 57.7%, and its break-even rate at 1.8R is 35.7%. Break-even sits inside the interval, so 50 trades cannot rule out a losing strategy even though the log made $1,160. If the win rate held at 44% and the payoff at 1.8R, the lower bound would clear 35.7% at 129 trades.
The interval has limits. It assumes trades are independent and that the true win rate stays constant. Strategies change with market conditions, and correlated trades (several positions opened on the same news) make the real uncertainty wider than the table shows. Average win and average loss carry sampling error of their own, so a confirmed win rate is necessary for a confirmed edge but not sufficient.
Losing streaks each win rate implies over 100 trades
A win rate also sets how long the losing runs get. The table gives the chance of at least one run of consecutive losses of each length somewhere in 100 independent trades.
| Win rate | 5+ losses in a row | 6+ in a row | 8+ in a row | 10+ in a row |
|---|---|---|---|---|
| 30% | 99.9% | 99.0% | 85.8% | 58.0% |
| 40% | 97.6% | 87.3% | 49.0% | 20.5% |
| 50% | 81.0% | 54.6% | 17.0% | 4.4% |
| 60% | 45.9% | 21.2% | 3.6% | 0.6% |
Method: exact recursion over the length of the current losing run across 100 trades, each independent with a fixed win rate; every cell agreed within 0.2 points with 200,000 simulated sequences per row. Correlated losses, such as several positions stopped out by the same news, make long runs more likely than the table shows.
A 40% win rate at 2R sits well above its 33.3% break-even and still carries a 49.0% chance of eight straight losses in 100 trades. Ten straight losses at 1% of the current balance each take 9.6% off the account. The losing streak answer covers what to check in the log when a run arrives.
Working with a small sample
- Report the count next to the percentage. "22 of 50, 44%" tells a reader how much weight the number can carry; "44%" alone does not.
- Add trades through testing, and label them. A backtest can add hundreds of trades, but backtested fills are not live fills, so keep the two samples apart.
- Avoid slicing a small log. Split 100 trades across four setups and each slice holds 25; 12 wins in 25 trades (48%) has a Wilson interval of 30.0% to 66.5%.
- Size for the uncertainty. While the interval still straddles break-even, the risk-per-trade answer and the position sizing explainer show how per-trade risk sets the damage a losing streak does.
Trading forex, CFDs and other leveraged products can result in losses that exceed your deposit, as the AlphaScala risk disclaimer states, and a win rate measured on past trades does not fix the outcome of the next one. This page is educational and is not personalised advice.
FAQ
What is a good win rate in trading?
A good win rate is one above the break-even rate for your own reward:risk after costs. A strategy whose average win is 1.8 times its average loss breaks even at 35.7%, so a 44% win rate at that payoff has positive expectancy before costs. A strategy whose average win is half its average loss needs more than 66.7%.
Is win rate the same as win/loss ratio?
No. Win rate divides wins by all closed trades; win/loss ratio divides wins by losses. 22 wins and 28 losses is a 44% win rate and a 0.79 win/loss ratio. Some traders use "win/loss ratio" for average win divided by average loss, so check the definition before comparing two figures.
How many trades do you need for a reliable win rate?
The count is set by how far the win rate sits above break-even. At 100 trades, an observed 50% still carries a 95% Wilson interval of 40.4% to 59.6%; at 300 trades it narrows to 44.4% to 55.6%. For any reward:risk from 0.5R to 3R before costs, a win rate 10 points above break-even takes 73 to 97 trades to confirm, and one 5 points above takes 289 to 385.
How many losing trades in a row are normal?
Over 100 independent trades, a strategy that wins 50% of the time has a 54.6% chance of at least one run of six straight losses. At a 40% win rate there is a 49.0% chance of eight losses in a row and a 20.5% chance of ten (AlphaScala calculation, 22 October 2026, exact recursion checked against 200,000 simulated sequences).
Should break-even trades count in the win rate?
Count them as trades but not as wins, which is how TradingView's "Percent profitable" metric treats "even" trades. Dropping them raises the figure: 21 wins, 27 losses and 2 scratches is a 42.0% win rate with the scratches counted and 43.75% without them.
Drafted with AI writing tools and published without an individual editor review. Figures, prices and dates may be out of date; check them against the primary sources before relying on them. Educational content only, not personalized financial advice.