How likely is it to succeed again after a short record of successes? The rule of succession calculator takes the successes and trials you have seen, and gives the probability of success next time, the probability that several future trials all succeed, the probability that the true rate exceeds a threshold and the uncertainty of the rate.
How the calculation works
- Enter the number of successes and the number of trials you have observed. The successes cannot exceed the trials.
- Laplace’s rule of succession says that the probability of success on the next trial is (s + 1) ÷ (n + 2). It acts as if you had seen one extra success and one extra failure.
- The calculator treats the unknown success rate with a uniform prior, so the posterior is a Beta(s + 1, n − s + 1) distribution. It uses this to find the chance that the next m trials all succeed and the chance that the true rate exceeds a threshold.
- The standard deviation of the posterior shows how uncertain the success rate still is.
Example
A new process has succeeded 18 times in 20 trials. The chance that the next trial succeeds is (18 + 1) ÷ (20 + 2) = 86.36 %. The chance that the next 5 trials all succeed is 19 × 20 × 21 × 22 × 23 ÷ (22 × 23 × 24 × 25 × 26) = 51.15 %. There is an 82.13 % probability that the true success rate is above 80 %.
Why it matters
Laplace used the rule to ask whether the sun will rise tomorrow, and it is still used to estimate small success or failure rates in testing, reliability work and A/B experiments, where observing zero failures does not mean the failure rate is zero.
Good to know
The rule assumes independent trials with a fixed success rate and a uniform prior, which is a strong assumption. Use it as a starting estimate and update it with other knowledge, especially when the data is very small.
Related tools
The Bayes theorem calculator shows how evidence updates a belief, and the binomial calculator gives probabilities when the rate is known.