Money Puck: The Effectiveness of Statistical Analysis in Building an NHL Team

The 2013 Collective Bargaining Agreement in the National Hockey League limits contracts offered to free agents in terms of length and variance in yearly salary. These changes have made finding undervalued free agents even more important to teams’ general managers. The purpose of this study is to evaluate players and teams with both traditional and advanced metrics to determine how players are valued in comparison to their impact on their team’s performance. A team’s winning percentage is hypothesized to be a function of shooting percentage and save percentage, as well as proxies for puck possession time, such as shots on goal per game, shots against per game, blocked shots, missed shots, and face-off percentage. It is also hypothesized that players with higher puck possession attributes will impact a team’s winning percentage to a greater extent than those with lower metrics, and so should be a key factor in determining how general managers use available salary money to improve their team. Based on data from NHL.com and stats.hockeyanalysis.com, we estimate team performance of all 30 NHL teams for each of the six previous seasons of play (2007-2013) as a function of puck possession proxies. We find that puck possession proxies significantly impact a team’s winning percentage and that free agents with higher performance metrics have a significantly greater impact on team performance as much as several less expensive players in cases that a team lacks depth