This project aimed to analyze the 2025 NFL Big Data Bowl tracking data to predict whether a play is a run-pass option (RPO) before the ball is snapped, and to identify the pre-snap “tells” that give RPOs away. An XGBoost model was trained on every frame from lineset to snap to produce a live RPO probability, and was tested with leave-one-week-out cross-validation against a logistic regression baseline and a SumerSports-style transformer. SHAP values were used to interpret the model, and the most influential features were combined into a league-wide tell sheet and team-by-team tell rankings that show which looks each offense uses most when running RPOs. Interactive play examples show the RPO probability changing live with motions and shifts. This document outlines all methods, data, and conclusions.

This project aimed to analyze the 2027 NFL Big Data Bowl data to determine whether NFL Scouting Combine results predict early-career success for wide receivers drafted from 2023 to 2025. Survival analysis, including Kaplan-Meier curves, a discrete-time hazard model, and Cox proportional hazards models, was used to measure how quickly receivers reached a successful season and whether combine measures add information beyond draft position. Additional season-by-season data from nflverse was also used in this analysis. This document outlines all methods, data, and conclusions.

This project aimed to analyze the 2022 NFL Big Data Bowl data to develop a method for identifying the "Best" kicker using data from the 2018-2020 seasons. Additional data from NFL weather and manually entered data were also used in this analysis. This document outlines all methods, data, and conclusions.