Effects of the Use of Sports Analytics and Team Attributes on Success in Regular Season of National Hockey League

In this paper, we study the effects of the use of sports analytics and team attributes on teams’ success in the  regular season of the National Hockey League. A team’s belief in analytics, the number of analytics staff,  and the number of professional staff hired are examined for the use of sports analytics. Some of the team attributes considered here are the average age of players in a team, payrolls of different positions (goalies,  defensemen, forwards), and numbers of the first-round draft picks in the previous three years. We shall examine the empirical data of 2014-2019 seasons. The team payroll is shown to be significantly positively  correlated with a team’s success in the regular season. It is interesting to see that teams scored 96 points or  more are very likely advancing to playoffs, whereas teams scored 92 points or less are very unlikely  advancing to playoffs. Four commonly used predictive modeling techniques (decision trees, random forests,  logistic regressions, and neural networks) are applied to the data for classifying teams into playoffs or no  playoffs. Random forests appear to be the best or as good as the other three techniques to yield the lowest validation misclassification error rate.

Keywords: Team payroll, decision trees, random forests, logistic regressions, neural networks.