Abstract
New directions in investigations which can be called numerical methods in mathematical statistics are being formed at this time. The necessity for an economical realization of algorithms to estimate the probabilistic characteristics on an electronic computer in the case of large quantities of samples dictates a criterion for the efficiency of estimate algorithms that is often different from that used in mathematical statistics. In this paper, criteria are formulated for the comparison of algorithms used in computer and programming practice, and examples are presented of the construction of quite efficient estimate algorithms on this basis. In particular, it is shown that an estimate in the form of the sample median can have advantages over the sample mean.