nflWAR: A Reproducible Method for Offensive Player Evaluation in Football

Abstract: Unlike other major professional sports, American football lacks comprehensive statistical ratings for player evaluation that are both reproducible and easily interpretable in terms of game outcomes. Existing methods for player evaluation in football depend heavily on proprietary data, are not reproducible, and lag behind those of other major sports. We present four contributions to the study of football statistics in order to address these issues. First, we develop the R package nflscrapR to provide easy access to publicly available play-by-play data from the National Football League (NFL) dating back to 2009. Second, we introduce a novel multinomial logistic regression approach for estimating the expected points for each play. Third, we use the expected points as input into a generalized additive model for estimating the win probability for each play. Fourth, we introduce our nflWAR framework, using multilevel models to isolate the contributions of individual offensive skill players, and providing estimates for their individual wins above replacement (WAR). We estimate the uncertainty in each player's WAR through a resampling approach specifically designed for football, and we present these results for the 2017 NFL season. We discuss how our reproducible WAR framework, built entirely on publicly available data, can be easily extended to estimate WAR for players at any position, provided that researchers have access to data specifying which players are on the field during each play. Finally, we discuss the potential implications of this work for NFL teams.

• The paper introduces nflWAR, a reproducible framework that leverages multilevel models combining expected points and win probability metrics to quantify offensive player impact.
• It utilizes the nflscrapR package to access NFL play-by-play data and applies multinomial logistic and generalized additive models for robust evaluation.
• The framework establishes replacement benchmarks and provides actionable insights to guide draft strategies, contract negotiations, and further research in sports analytics.

A Comprehensive Framework for NFL Player Evaluation

The paper by Yurko, Ventura, and Horowitz addresses the lack of comprehensive, reproducible statistical ratings for player evaluation in American football, particularly within the National Football League (NFL). Their work bridges the gap that football analytics currently suffers when compared to other major sports such as baseball and basketball, which have well-established systems for assessing player value related directly to game outcomes like points or wins. The authors propose a multifaceted approach that lays the foundation for a reproducible framework for evaluating NFL players, particularly offensive skill positions, with potential extensions for broader applications.

Key Contributions

The key contributions of the paper, encapsulated within the framework termed "nflWAR," include:

1. Data Access through nflscrapR: The authors developed an R package, nflscrapR, which provides access to publicly available NFL play-by-play data. This tool serves as a resource for further research and analysis, addressing the scarcity of accessible data that hampers public research in football.
2. Expected Points (EP) Model: The paper introduces a multinomial logistic regression model to estimate the expected points model for each play. This statistical approach appropriately models the "next score" response variable, marking an advancement over previously opaque methodologies.
3. Win Probability (WP) Model: This model uses a generalized additive model (GAM), incorporating expected points and other covariates, to estimate the win probability for each play. This approach does not account for team strength, aligning estimations to individual effects rather than team-based performance factors.
4. WAR Framework: The nflWAR framework employs multilevel models that capture the contribution of individual offensive skill players to the team's success, quantified as Wins Above Replacement (WAR). The calculations are robust to positional roles and account for contributions beyond raw statistics, including situational context and teammate quality.

Notable Results and Methodologies

The results presented in the paper focus on the 2017 NFL season, providing empirical validation of the nflWAR framework's ability to delineate player performance. The authors showcase individual WAR metrics while also partitioning contributions into components like passing, rushing, and yards after catch (YAC). For instance, they highlight how different styles of play among quarterbacks (QBs) and running backs (RBs) are translated into their WAR scores, showcasing detailed breakdowns and insights such as Tom Brady's high passing efficiency against his limited rushing value.

One significant numerical result includes the establishment of replacement level benchmarks for players' evaluation. This allows for comparisons against a "freely available" player standard across different positions, a well-discussed point for embracing economic-like metrics into sports analytics.

Implications and Future Research Directions

The theoretical implications of this research are profound, as the structured methodological approach could serve as a template for future player evaluation frameworks, extending into more granular applications such as defensive player assessment. The practical implications suggest substantial potential for NFL teams in optimizing player evaluation, draft strategies, and contract negotiations through an objective, data-driven lens.

A key area for future development is the integration of more granular data, such as player-tracking or team positional data, which would further refine individual contributions and improve estimation accuracy for positions like offensive linemen. Additionally, the ongoing development of the R package and public data repositories presents opportunities for continuous improvement and community engagement in driving research advancements.

In summary, the paper by Yurko, Ventura, and Horowitz establishes a reproducible and statistically sound method for offensive player evaluation in the NFL, contributing significantly to the field of sports analytics. The framework's scalability and adaptability mark an essential step toward comprehensive football analytics, with implications that could reshape team decision-making processes and enrich sports statistical methodologies.