This model predicts a player’s average fantasy performance over their next four games based on their most recent seven-game history. It uses a Bayesian LSTM with attention to capture sequential trends in player statistics, incorporating both game-by-game performance and contextual embeddings for player position and years of experience.
The model outputs projections for the standard Yahoo 9-category fantasy basketball metrics (points, rebounds, assists, steals, blocks, turnovers, FG%, FT%, and 3PM), with shooting statistics represented using made and attempted shots to reflect both efficiency and volume.
Note: This model is still under active development. Prediction results are for reference and exploratory analysis only.
This app includes an intent-aware AI analysis layer to help translate prediction results into actionable Yahoo Fantasy basketball advice.
A detailed project report covering the modeling approach, data pipeline, uncertainty estimation, and deployment architecture is available below:
👉 View Full Project Presentation