This paper is dedicated to discussing methods of statistical modeling the outcomes of sport events and, particularly, matches with continuous time. We propose a simulation-based approach to predicting the outcome of a match, somehow medium between pure statistical methods and agent simulation of individual players. An example of retrospective prediction is given.
A computer program intended for detecting regularities and forecasting the results of men’s 100 m of the World Cup -2015 in track and field athletics is developed. The neural network based on the results of the previous World Cups and the Olympic Games lies in the heart of the program. The error of forecasts is no more than three percent. Besides forecasts the program allows to estimate the influence of the parameters’ change characterizing athletes on heir sports results, and also to select optimum combinations of these parameters for each athlete. The dependences of probability of a victory in the World Cup from age, weight, muscular weight, starting reaction and other parameters are received for some well-known athletes. By research of neuronetwork mathematical model recommendations about improvement of productivity of well-known athletes are developed: Useyn Bolte, Tyson Gay, Christoff Lemetr, Nessa Carter, Johan Blake and Justin Getlin.