Pravděpodobnostní modely a metody pro analýzu sportovních výsledků

Abstract

The statistical modelling on a sports field is in the process of a great development. Sports clubs have gradually changed into commercial companies and thus many possibilities how to take advantages of statistical instruments have emerged. Undoubtedly, the most popular task is the prediction of possible future results. The first part of this thesis summarizes the most interesting methods and types of solutions that were developed primarily for football matches. In this case the Poissons' regression models are mainly used. However, when the influences of time are taken into consideration, these models become more difficult and complicated for computations. The second part deals with the statistical survival analysis and random point processes and tries to find its utilization and connection with the predictions of the results of sports matches. On the basis of the Cox' regression model, a simple model for ice-hockey matches is derived here. This model estimates offensive and defensive forces of every team and other influences, for instance power plays and home environment. It describes the match as a random point process, where the time of scoring represents its random points. The model is applied on the data obtained from the first thirteen rounds of the Czech National Ice-hockey League and the received...