356,371 unique french races in a database great to feed new frontier models

Hugging Face released Huggy Database, the largest open dataset of French horse racing with 356,371 unique races from 1996 to 2026, covering 30 years of daily data including ~5M runners, jockey/trainer/owner details, odds, and payouts. The dataset is clean, deduplicated, and ML-ready, designed for tasks like winner prediction, odds efficiency research, and graph ML on racing networks.

Hi HF community Just released Huggy Database - the largest open dataset of French horse racing ever published on the Hub. What’s inside 356,371 unique races 1996 → 2026, 30 years of complete daily coverage ~5M runners with jockey / trainer / owner / weight / odds / finishing position Payouts PMU / PMH — Simple, Couplé, Trio, Quinté+, etc. - Clean, deduplicated, normalized accents/case-insensitive matching on 40K+ actors - ML-ready CSV + relational structure Links Use cases - Tabular classification winner prediction - Odds efficiency / market inefficiency research - Long-horizon sports analytics 26 years of data - Graph ML on jockey/trainer/horse networks Feedback welcome - What columns would you add? - Interested in a HF-hosted Parquet mirror? - Any specific ML baseline you’d like to see?