Predicting NFL Rushing Yards with Contextual Factors

This study develops and evaluates predictive models for NFL running back rushing yards by incorporating

contextual and matchup-specific features beyond traditional statistical averages. Using a dataset

of 2,861 player-game observations from the 2018-2024 NFL seasons, we compared six

modeling approaches: baseline linear regression, enhanced linear regression,

Ridge regression, Lasso regression, Random Forest, and Gradient Boosting. The enhanced linear

regression model achieved the best performance with a validation set RMSE of 16.59 yards

and R² of 0.532, representing a 2.7% improvement over the baseline model. Key findings indicate

that contextual factors including precipitation, home field advantage, and opponent defensive