PILOT EVALUATION OF THE WEIGHTED OFFENSIVE PRODUCTIVITY RATING IN HOCKEY USING LONGITUDINAL DATA AND LOGISTIC REGRESSION
Abstract
The purpose of this study was to evaluate an analog to the baseball statistic Weighted On-Base Average (wOBA), which provides an estimate of a baseball player’s offensive production, in the context of hockey. In this model, a hockey player’s offensive skillset is determined by passing, zone-transition, and shooting, which together create a Weighted Offensive Productivity Rating (wOPR). The contribution of these components were estimated using shot assist, shot attempt, zone entry, and zone exit data, with linear weights produced separately for forwards and defensemen. Logistic regression estimations were used to analyze whether the wOPR is a better predictor of game success compared to a more traditional measure such as a player’s total points in NCAA Division I Hockey. Each prediction equation consisted of the performance metric, the quality of competition, and a quality of teammates’ metric for any given player in a given game. The wOPR at the team level in 5 on 5 play is a statistically significant predictor of game results. The wOPR appears to perform better as an indicator of play quality in instances where data availability about individual players may be limited, e.g. new players or early season play. Preliminary analysis supports the adoption of the model in practice.
Collections
- OU - Theses [2188]