Projecting the Future Individual Contributions of NHL Players

Abstract

Professional sports are a multibillion-dollar industry with millions of people invested in the outcomes of games and seasons. Owners, management, and fans sit on the edges of their seats wondering what will happen next. Lots of work has been done forecasting success at the team level across a variety of sports, but player level predictions are less common. Predictive work related to the NHL is even rarer. This thesis explores the ability to predict NHL player performance in a given season using publicly available information via statistical learning methods. Data featured in the analysis includes play-by-play and shift information, box score statistics, a variety of composite and catch-all statistics, injury information, and player biographical information. Data was compiled and analyzed to find meaningful relationships between past and future performance. The results of the analysis found the most predictive values in . season’s raw numbers can be supplemented with more information to improve predictive power.