Estimating the Strength of the Impact of Rushing Attempt in NFL Game Outcomes
Xupin Zhang *
University of Rochester, Rochester, NY 14620, USA.
Benjamin Rollins
Rochester Institute of Technology, Rochester, NY 14623, USA.
Necla Gunduz
Department of Statistics, Gazi University, Ankara, Turkey.
Ernest Fokoue
Rochester Institute of Technology, Rochester, NY 14623, USA.
*Author to whom correspondence should be addressed.
Abstract
In this paper, we use estimators of variable importance from the ensemble learning technique of random forest to consistently discover and extract the knowledge that Rush Attempt is strongly related with winning football games in the NFL. Almost all researchers before us have consistently made claims of the impact/influence other statistics in the outcomes of NFL games, with Third Down Conversion Percentage and Takeaways almost universally considered as having the greatest impacts in game outcomes. Rushing as a factor of NFL success has also been mentioned, but mostly in terms of number of rushing yards per game. The novelty in this present work lies in the fact that not only do we discover Rush Attempt differential to be the strongest and most dominant variable, but we also establish its dominance throughout the years, namely with 14 seasons worth of NFL games data providing firm evidence of the ubiquitous appearance of Rush Attempt at the root of every classification tree.
Keywords: NFL, rush attempt, classification trees, and random forests