NBA "What-If" Analyzer and Play Editor
A full-stack interactive web application with a React frontend, Flask API, and an ONNX-deployed MLP neural network to model real-time NBA win probability and simulate hypothetical game-state edits.
Existing win-probability graphs are static post-game visualizers that do not allow users or analysts to test counterfactual scenarios (e.g. 'what if the missed 3-pointer went in?').
Combined a React + Vite user interface with a Flask backend powered by nba_api data and an ONNX MLP neural network trained on historical play-by-play momentum curves to dynamically recompute win probability graphs when users edit historical game events.
Built a real-time interactive sandbox where analysts can alter play outcomes, view live win probability shifts, and compare prediction market odds from Kalshi and The-Odds-API.