It might be a simple question, but still difficult for me.
What would be a good way to estimate probability from frequency of occurrence if we have finite number of observations?
Example. Some experiment gives $0$ or $1$, we count experiments where we get $0$ and estimate it probability of $0$ as $p = N_0/N$, where $N$ overall number of experiments.
However in the definition true probability value can be computed having infinite number of experiments, which is impractical.
Is there another(maybe better) way to compute probability in the mentioned example. Or how many experiments $N$ need to be executed so we have some confidence in the estimation? (Sorry if it doesn't make much sense)