Search

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.

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 different trajectories of the aging curve. It was also found that midfielders and defenders have a longer near-peak level of performance compared to forwards and goalkeepers.