ATP Tennis Match Prediction Model

This repository contains code, data, and saved models for predicting ATP match outcomes. The pipeline builds chronological Elo and serve/return features, then trains calibrated probability models. The 2026 season is held out for evaluation.

scripts/tennis_model/: feature generation, training, and walk-forward evaluation
data/: source and processed match data; yearly odds workbooks are in data/sportsbook/
models/: saved model and player-state artifacts
output/: audits and backtest results
web/: dashboard source
web/predictions/kalshi/: Kalshi connection and live order routes

See PROJECT_DETAILS.md for methods, results, and run commands. Hugging Face hosts these files as a model repository; it does not run the dashboards.

Downloads last month: -; Downloads are not tracked for this model. How to track