Player impact measures for scoring in ice hockey

• A commonly used method to evaluate player performance is to attribute values to the different actions that players perform and sum up these values every time a player performs these actions. In ice hockey, such metrics include the number of goals, assists, points, plus-minus statistics and recently Corsi and Fenwick. However, these metrics do not capture the context of player actions and the impact they have on the outcome of later actions. Therefore, recent works have introduced more advanced metrics that take into account the context of the actions and perform look-ahead. The use of look-ahead is particularly valuable in low-scoring sports such as ice hockey. In this paper, we first extend a recent approach based on reinforcement learning for measuring a player's impact on a team's scoring. Second, using NHL play-by-play data for several regular seasons, we analyze and compare these and other traditional measures of player impact. Third, we introduce notions of streaks and show that these may provide information about good players, but do not provide a good predictor for the impact that a player will have the next game. Finally, streaks are compared for different player categories, highlighting differences between player positions and correlations with player salaries.

Subject headings

NATURVETENSKAP  -- Data- och informationsvetenskap -- Datavetenskap (hsv//swe)

NATURAL SCIENCES  -- Computer and Information Sciences -- Computer Sciences (hsv//eng)

Publication and Content Type

ref (subject category)

kon (subject category)