ADS Capstone Chronicles Revised

1

NFL Teams Should Focus On Passing Caleb McCurdy

UEWang Applied Data Science Master’s Program Shiley Marcos School of Engineering / University of San Diego yuewang@sandiego.edu 1 Introduction One of the greatest aspects of the National Football League (NFL) is there are many different avenues teams can take en route to the ultimate goal of hoisting the Lombardi Trophy. Some general managers place greater weight on draft picks than others, which results in willingness to make these assets available via trade. One of the benefits of building out the roster with an emphasis on the draft is the youth it brings to the team. Additionally, the average contracts are lower for these players before hitting their first official free agency or receiving a contract extension. However, other teams prefer to sacrifice some of these assets to build with players that have proven their abilities at the professional level. Although teams have various philosophies on how to build out their roster as well as how to utilize the players’ talents, we will analyze the predictive power of different in-game factors to determine which have the greatest impact on winning. We will utilize regression and classification models to evaluate how closely we can predict the winner of games based on the most important, non-scoring factors. 2 Background In the current state of the NFL, teams have various schematic philosophies. As a result, players are seen as more valuable to certain teams than others as they match the capabilities of that individual scheme. Though there is no one singular way to win games in the NFL, made

Applied Data Science Master’s Program Shiley Marcos School of Engineering / University of San Diego calebmccurdy@sandiego.edu consistently employ differing roster building philosophies in hopes of gaining a competitive advantage over their competition. Though there are generally agreed upon needs, such as a competent quarterback and coaching that maximizes talent, teams fill out the rest of their rosters in various ways. As the NFL evolves, we investigate whether building a roster and scheme focused primarily on the passing game leads to greater success as well as higher scoring outputs. Through exploratory analysis, modeling, and evaluation, we demonstrate scoring is, in fact, more greatly affected by an increase in passing efficiency than by increasing rushing efficiency. Additionally, there are specific in-game factors more correlated with winning, as we are able to build a logistic regression model with over 82% accuracy of win predictions on unseen data, even while excluding direct scoring factors. There is strong evidence to support the claim teams can expect greater success through maximizing efficiency in the passing game. Future studies were recommended to address overfitting and explore incorporating additional data sources for further model refinement. Overall, this analysis offers valuable insights into predictive modeling for NFL game outcomes and scoring, providing practical applications for NFL sports analytics and decision-making processes. KEYWORDS free agency; NFL draft; roster building; competitive imbalance; statistical analysis ABSTRACT National Football League (NFL) teams

28