Empirical vs theoretical probability