|
Abstract:
|
Beane (Lemire 2015) identified injury prevention as the next frontier in sports analytics. In this study, I demonstrate that statistics is essential by examining re-injury prevention, through testing whether NCAA coaches and players, facing the need to limit repetitions as key players recover from injury, allocate those repetitions in a manner that maximizes win probability. I calculate in-game win probabilities using the methodology of Stern (1991) and Winston (2009), NCAA historical point spread data from Covers.com, and NCAA expected points of each down-distance-position state from Knowlton (2015). As a test case, I consider Brigham Young University's use of dual-threat quarterback Taysom Hill in his 2016 return from a prior-year Lisfranc injury. I extract play-by-play data from BYUCougars.com using R's rvest package and regular expression functions. I determine the situations in which Hill's rushing, based on his yards per carry prior to the game on standard and passing downs, maximizes changes in win probability.
|
