Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let's say we have an unfair dice with the initial (estimated) probability of rolling a specific number = 1/6. How does probability estimation update when getting this specific number on every next experiment? How does probability change when getting other number (on failure)?


share|cite|improve this question
up vote 3 down vote accepted

You should not start with a point estimate but with a prior distribution, which we might call $f(p)$ where $p$ is the unknown probability.

For example, suppose you are trying to roll a 6. You might get $x$ successes and $y$ failures. The likelihood of this is proportional to $p^x (1-p)^y$. So your posterior distribution is

$$f(p|x,y) = \frac{p^x (1-p)^y f(p)}{\int p^x (1-p)^y f(p) \, dp}$$ with the integral taken over the possible range of $p$, which is typically $[0,1]$ for a probability.

If you start with a prior distribution which is $f(p)=\text{Beta}(\alpha,\beta)$ then you will get a posterior distribution of $f(p|x,y)=\text{Beta}(\alpha+x,\beta+y)$: this is called a conjugate prior and assuming it makes the mathematics easier. The expected value of a Beta distribution $\text{Beta}(\alpha,\beta)$ is $\frac{\alpha}{\alpha+\beta}$, so you might decide to start with $\beta = 5 \alpha$, and set $\alpha$ according to how confident you are at the start that $p$ is close to $\frac{1}{6}$. This would make the expectation value of your posterior distribution $\frac{\alpha+x}{\alpha+x+\beta+y}$.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.