Random number generators are used in algorithms for outcome generating in games of chance. We present the six Knuth's tests. They check whether probabilities of generated outcomes match the theoretical probabilities. Each one of them reduces to Pearson's $\chi^2$ test, which has the $\chi^2$ distribution for large random samples. That is how we calculate $p$-values based on which we evaluate the fairness of outcome generators. Despite Pearson's $\chi^2$ test being widely used, it is complex, since a big random sample is needed for its accuracy. Therefore the recently published alternative $\chi^2$ test is also presented in the thesis.
|