Search

Understanding the role of a particular player, or set of players, in a team is an important tool for players, scouts, and managers, as it can improve training, game adjustments and team construction. In this paper, we propose a probabilistic method for quantifying player roles in ice hockey that allows for a player to belong to different roles with some probability. Using data from the 2021–2022 NHL season, we analyze and group players into clusters. We show the use of the clusters by an examination of the relationship between player role and contract, as well as between role distribution in a team and team success in terms of reaching the playoffs.

This paper explores the significance of special teams, particularlypowerplay, in ice hockey. Despite the commonly held perception oftheir importance, little research has examined the impact of powerplayand penalty kill performance on overall team success. The paper usesseveral seasons of NHL data to characterize goal-scoring and manpoweropportunities, and perform analysis from several perspectives. The resultsindicate that individual even strength goals and powerplay goalshave similar value, but the larger share of even strength goals scoredover a season makes even strength play a more important contributorto team success. The paper also finds a high correlation between teamsthat perform above/below average during even strength and powerplay.This study provides valuable insights into the dynamics of ice hockeygameplay and the role of special teams in determining team success.

To capture that not all goals are of the same importance, a new performance metric called the Game Points Importance Value (GPIV) was recently proposed. While this metric takes into account the expected impact that a goal has on the outcome of a game based on the context when the goal was scored, it relies on a relatively fine-grained state space. To address this problem, this paper presents simplified and more practical variations of the GPIV metric. Motivated by our analysis of the relative importance of different dimensions of the state space, we present two metrics that capture the most important component(s) of GPIV. Our evaluation shows that the metrics are relatively stable and capture most of the relative differences between GPIV and traditional metrics (e.g., goals, assist, points, and +/-). These results suggest that these simple and practical metrics are intuitive, capture most of the desirable variations that GPIV captures, and that the value of a goal can be well estimated using GPIV data based on historic data.

Within sports, the relationship between performance and age is deeply intertwined and is of interest to various roles within the sport. Despite its status as one of the largest sports in the world, there is a limited amount of previous work in the context of football and the leagues investigated has emphasized the highest ranked leagues. As a result, leagues with a lower ranking as well as lower tiers remain largely unexplored. Furthermore, the performance metrics used previously were primarily focused on summarized statistics from each match rather than the events themselves. Using more than one performance metric was also rare. In this thesis, different performance metrics were first examined and computed and then used to evaluate player performance on data from the three highest tiers in Swedish male football hierarchy. The performance metrics used were Expected Threat, Post-shot Expected Goals, as well as offensive and defensive values from the Valuing Actions by Estimating Probabilities framework. The importance of the metrics varied by position and should be weighted accordingly for optimal performance. A weighted sum of the metrics was then used as the target, which was modeled by the delta method, polynomial regression and generalized additive models. The features used were either age, age and player, or an interaction between age and age-related decay. Among the selected models, the delta method was not suitable for this type of data but both polynomial regression and generalized additive models showed promise in peak age estimation. The results indicated that the peak age varies by position but also on the method of estimation. For forwards, midfielders and defenders the peak performance was most likely to occur between the ages of 25 and 27 while the apex among goalkeepers frequently occurred in the age range of 25 to 29. The polynomial regression implied a later peak in general than the generalized additive model and the latter also allowed for differe