Shot Analysis of NHL Players and Its Application in the Training Process

This diploma thesis focuses on the application of cluster analysis to NHL shot data from the 2023/2024 season. The main objective is to identify the optimal way to group shots and players based on their spatial, technical, and game-related characteristics, and to interpret the resulting clusters generated by the most suitable method. The secondary objectives include a comparison of seven clustering methods (e.g., k-means, Ward’s method, single linkage) and various distance metrics in terms of their ability to form statistically and practically meaningful groups. To evaluate the quality of the results, three standard internal validation metrics are used: the silhouette coefficient, Calinski-Harabasz index, and Davies-Bouldin index. The analysis is divided into three parts: clustering of shots based on their position on the ice, based on shot type and outcome, and clustering of players according to their physical and game characteristics. The results demonstrate that cluster analysis is an effective tool for identifying typical shooting patterns and player profiles. The thesis formulates specific recommendations for incorporating these insights into training practice and highlights the potential of data analysis to support decision-making in ice hockey.