I n this project, I developed a predictive model to classify NFL plays as run-left, run-right, or run-center based on pre-snap player movement. Using the NFL Kaggle dataset, I focused on identifying the fastest-moving player before the snap and extracted key motion-based features such as speed, acceleration, direction, orientation, and velocity. By engineering features like forward and sideways velocity, orientation sine and cosine transformations, and relative positional differences, I aimed to capture the nuances of player movement leading up to the play.

To build the predictive model, I implemented a Random Forest classifier, leveraging its ability to handle complex, non-linear interactions between movement-based features. The model was trained to recognize patterns in pre-snap motion and predict the intended direction of the play with a high degree of accuracy. Through iterative feature selection and hyperparameter tuning, I optimized the model’s performance, ensuring that it generalizes well across different teams and formations.

To enhance interpretability, I developed a Matplotlib-based visualization tool that animates plays frame by frame, allowing users to see the movement of players leading up to the snap. This visualization highlights how the fastest-moving player influences play direction and serves as a valuable tool for analyzing trends in offensive strategy. This project showcases my ability to apply machine learning to sports analytics, combining data-driven insights with dynamic visual storytelling. If you want to check out the project, here is the Github Repository.