Cross-Sport ML Analytics Platform
Overview
A full-scale machine learning analytics platform that explores cross-sport performance patterns. Built a unified 'Player Impact Score' that works across NBA and soccer, discovered cross-sport archetypes (e.g., 'stretch-4' maps to 'inverted fullback'), and developed breakout prediction models. Features an interactive Streamlit dashboard with explorer tools for deep analysis.
The Problem
Sports analytics is siloed — NBA analysts rarely share methods with soccer analysts. Can ML models and player evaluation frameworks actually transfer across fundamentally different sports?
My Approach
Built a unified feature space mapping NBA and soccer stats into 6 abstract dimensions (scoring, playmaking, defense, efficiency, volume, versatility). Trained impact models, player clustering, and breakout predictors that work across both sports. Used SHAP for interpretability and built an interactive Streamlit dashboard for exploration.