Home › Calculators › Win Rate Confidence Interval Calculator

Win Rate Confidence Interval Calculator

The win rate in a sample is an estimate, not an exact value. If predictions have some true probability of winning, then over a hundred predictions the share of wins could turn out to be 54% or 66% purely by chance. The calculator shows which range of probabilities is consistent with your data.

Win rate confidence interval

How precisely the share of wins in a sample describes the true probability of winning.

Sample win rate—
Lower bound—
Upper bound—
Examples:

Wilson interval for the share of wins. It describes only the statistical uncertainty of that share: it does not predict the next run, does not show profitability and does not check whether the prediction history is complete.

What a confidence interval shows

Take 60 wins out of 100 predictions. The share of wins in the sample is 60%, and the 95% Wilson interval is 50.2% to 69.1%.

The 95% confidence level refers to the method, not to a specific interval. If you collected samples of 100 predictions many times and built an interval for each in the same way, about 95% of those intervals would contain the true probability of winning. For a single sample it means: values inside the interval agree well with the data, values outside it agree noticeably worse.

How the interval width depends on the sample

PredictionsWonSample share95% interval
2000%0–16.1%
201365%43.3–81.9%
2020100%83.9–100%
503060%46.2–72.4%
1006060%50.2–69.1%
30018060%54.4–65.4%
1,00060060%56.9–63.0%
5,0003,00060%58.6–61.3%

The width of the interval shrinks roughly in proportion to 1/√n: to halve the interval you need about four times as many predictions.

The extreme cases make sense too. 0 wins out of 20 does not mean a zero probability — the data are compatible with a win probability of up to 16.1%. And 20 out of 20 is compatible with a probability of 83.9%.

The Wilson method

Let n be the number of predictions, p̂ the share of wins in the sample, and z the quantile of the normal distribution for the chosen confidence level: 1.645 for 90%, 1.960 for 95% and 2.576 for 99%. Then:

centre = (p̂ + z² / 2n) / (1 + z² / n)

half-width = z × √(p̂(1 − p̂) / n + z² / 4n²) / (1 + z² / n)

The interval is centre ± half-width.

The simpler formula p̂ ± z × √(p̂(1 − p̂) / n), the Wald interval, performs poorly on small samples and with shares close to 0% or 100%. With 0 or n wins it gives an interval of zero width, and in other cases it can go beyond 0–100% and cover the true value less often than the stated level. The Wilson interval is almost free of these flaws, which is also why its bounds are not symmetrical around the observed share: near the edges of the scale it shifts towards the middle.

Assumptions of the calculation

What a confidence interval does not show

Frequently asked questions

What does a 95% confidence level mean?

It is a property of the method: when samples are repeated many times, about 95% of intervals built this way contain the true probability of winning. It is not the probability that the true value lies inside one particular interval.

Why is the interval not symmetrical?

The share of wins cannot be less than 0% or more than 100%. The Wilson interval takes this into account and shifts towards the middle near the edges of the scale, so at 90% wins the distance to the upper bound is smaller than to the lower one.

What if wins are 0 or 100%?

The calculator handles both cases correctly: one of the bounds equals 0% or 100%, and the other shows how far from it the true probability could be with such a sample.

Can the interval tell you whether the predictions are profitable?

No. The interval describes only the share of wins. To assess profitability you need the odds and result of each bet, as well as the full history of predictions published in advance.

How many predictions are needed for the interval to become narrow?

With a share of wins of about 50–60%, an interval of roughly ±10 percentage points is reached at 100 predictions, ±3 pp — at about 1,000, ±1.4 pp — at 5,000.

This calculator is for educational purposes. Sports betting involves the risk of financial loss.

More predictions in your account

The full selection of matches with odds from 2.00.

Get access

More calculators