You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
GCI World 2026 April - NFL Draft Prediction. Binary classification (AUC) solution using a LightGBM/XGBoost/CatBoost stacking ensemble with Optuna tuning. Matsuo-Iwasawa Lab, University of Tokyo.
Completed an analysis on the relationship between the combine performance and future league performance of 941 NFL players drafted between 2015 and 2017 using R. Proved that the combine performance of NFL players has little to no correlation on future player performance using linear regression. My goal was to show that the NFL Combine should not…
Predicting NFL Draft selection probability from Combine metrics using a tuned LightGBM + XGBoost ensemble. Features position-normalised z-scores, smoothed school target encoding, and Optuna hyperparameter tuning. Built for GCI World 2026 April In-Class Competition. Public AUC: 0.82983.
Can we predict an NFL rookie's success using only raw Combine traits? This project evolves my academic coursework into a fully modular, object-oriented ML pipeline. This model mathematically grades the 2026 draft class - leveling up my enterprise architecture skills while giving me an excuse to over-analyze the Chicago Bears' draft picks.
Predictive model and code used to determine the relationship between variables collected from each game and the number of wins for a given team in 2020.