HoopMind: AI System for Real-Time Basketball Strategy

A new AI system called HoopMind has been developed to provide professional-level basketball analytics using publicly available data. The system integrates five distinct public data sources, including shot locations, play-by-play feeds, matchup tracking, and player biometrics, into a comprehensive dataset of over 4 million shots across 21 seasons. This extensive data fusion allows for detailed analysis of game dynamics. HoopMind models a half-court possession as a sequential game, utilizing ShotNet, an embedding multilayer perceptron, to predict shot values. This model outperforms baseline methods and provides well-calibrated probabilities. An expectimax search with branch-and-bound pruning then solves the offensive decision tree in real-time, with all training performed offline to keep the online system lightweight. The system offers both a scouting planner and a playable simulator, accessible within a single browser page. This allows coaches and analysts to prepare for opponents with data-driven insights, bridging the gap between traditional intuition-based coaching and advanced professional analytics tools.