Generalized Linear Mixed Models for Putting Performance on the PGA Tour
MetadataShow full item record
In golf, putting is considered to be the single most important aspect of the game. With approximately 40 percent of total strokes having been putts in 2013 on the PGA TOUR, it is obvious that putting performance is a significant aspect of a player's overall performance. Analysts are always trying to find new ways to measure a player's putting performance on the golf course. For many years, measuring putting performance has been restricted to simply measuring putts per round or putts per green. In this paper, we concentrate on quantifying the various factors that contribute to putting performance. While the probability of making a putt as a function of distance has been modeled in past research, a mixed effects model has not yet been used to incorporate additional variables representing random effects such as individual or season-to-season variability. A series of generalized linear models and generalized linear mixed models using the logit link function were used in this analysis. The response variable used was whether or not the putt finished in the hole, making the logit link function ideal for this analysis. The BIC and McFadden's R2 model selection criteria were used for model comparison. The best model consisted of fixed effects; Distance and Putt.For, and random effects; Year within Player and Hole within the Course. An analysis of the best linear unbiased predictors (BLUPs) also provides insight into the conditions for which putting performance is at its best. Additionally, a generalized linear mixed model was _t for the 2014 season and the BLUPs were used as a ranking system for putting performance. The results were compared to the rankings provided by strokes gained putting and the ranking systems showed moderate consistency.