Applications of multivariate statistical methods in ice hockey

The diploma thesis focuses on analysing Czech Tipsport Ice Hockey Extraliga players’ data with a main goal of gaining knowledge helping with characterizing ice hockey players and evaluating their influence on team’s play. First two analyses use forwards’ even-strength data from season 2020/2021, the third analysis focuses on seasons since 2014/2015. First partial goal of this thesis is to make large number of players’ metrics simpler and create smaller number of independent factors using factor analysis. Final model uses thirteen well-interpretable factors which are described and named. Second partial goal is to find homogeneous groups of forwards by applying cluster analysis. Eighteen clusters found well characterise various playing styles and despite larger number of clusters the results are usable, especially to identify players with similar style. Third partial goal is to estimate players’ influence on teams play using Bayesian approach to regression analysis and dummy observations to formulate prior information. Models use shot activity and goal production as target variables. Estimates of parameters of regression RAPM model describe players’ influence on the game and quantify influence of other aspects of game. Posterior distributions of selected parameters are visualized, also distribution of point estimates of player parameters throughout all seasons are analysed. Analysis in this thesis was performed in RStudio using programming language R.