This is a review of the Weighted Least Squares Regression Model research with NHL data conducted by Brian Macdonald.
NBA analysts and teams use the adjusted plus-minus (APM) stat to determine players’ contributions to the offense and defense. One strength of the APM is that each player’s score is not dependent on his teammates’ scores which is an improvement over the traditional plus/minus stat in which players’ scores were dependent on their teammates. This APM model can be adjusted for use in the NHL. The issue becomes that hockey games are not always played at even strength. Teams can be down one, or even two players during a penalty. Including these situations in the APM would be unfair to those players on the ice as it would lower their ratings, in some cases quite drastically.
So, a new model needed to be developed. In fact, two models were necessary – one for special team situations and one for even strength situations.
The model for even strength situations includes both an offensive and defensive statistic with the exception of the goalie who will have only a defensive statistic. The information will be inputted on a shift by shift basis with a shift being a period of time when no player substitutions are made. The offensive stat is based on goals scored by the team and the defensive stat is based on goals scored against the team. Then the model is adjusted to take into account which zone the faceoff takes place in.
The model for special team situations will account for both power play and shorthanded situations. Therefore, four stats will be needed – goals scored for and against during power plays, and goals scored for and against while shorthanded.
To get the total APM the scores from the two models are added together.
Using this model would give a more accurate picture of the contributions each player makes to their team. Analysts can compare players across the league in terms of total plus/minus, offensive plus/minus, defensive plus/minus, even strength plus/ minus and special team plus/minus. This will allow analysts to gain a more rounded picture of how players compare with others in the league.
Teams can look at the individual stats when they are looking at trading players. If they know that their team is weak in one area, they will be able to look at the stats of the available players to determine which player would be most likely to improve their team in that area. It would also help coaches work out training plans for improvement in these specific areas.
How mature is your team’s analytics program? Take the Sports Analytics Maturity Assessment.
Learn about the Groundbreaking Sports Analytics Model coaches and sports analysts are talking about!
Learn all about Sports Analytics here.