Huggy Engine has published the Huggy Database on Hugging Face, described as the largest open dataset of French horse racing available. The release provides 30 years of complete daily coverage from 1996 to 2026 for use in machine learning applications.

  • Contains 356,371 unique races with approximately 5 million runners.
  • Includes detailed attributes such as jockey, trainer, owner, weight, odds, and finishing position.
  • Covers various payout types including PMU, PMH, Simple, Couplé, Trio, and Quinté+.
  • Data is clean, deduplicated, and normalized with case-insensitive matching for over 40,000 actors.
  • Structured as ML-ready CSV files with a relational schema.

The dataset supports use cases such as tabular classification for winner prediction, odds efficiency research, long-horizon sports analytics, and graph machine learning on jockey, trainer, and horse networks.