The Huggy AI Challenge invites participants to predict the exact top-5 finishers of the Quinté+ horse race held on 23 August 2026. The goal is to determine if an AI model can out-guess the parimutuel market using an open protocol and public metric. Any AI method is eligible, including GPT, Claude, Gemini, Mistral, DeepSeek, Qwen running locally, in-house models, or custom methods.
The competition consists of two phases. Phase 1 runs from now until 21 August, serving as a rehearsal on historical races. The leaderboard during this phase is scored against past Quinté+ races held out of the training set, allowing participants to build, debug, and calibrate their models without ranking anyone on the real race. Phase 2 takes place on 23 August. On 21 August 2026 at 10:00 Paris time, the runners for the 23 August race are published in the Data tab with detailed history including form, jockey and trainer statistics, shoeing, kilometric reduction, and performance metrics. Submissions close before the official start time of the race. After the official arrival, the solution file is replaced by the real result, and the final leaderboard is computed on that race only. Note that Phase 1 submissions do not carry over; you must submit again between 21 and 23 August to appear on the final leaderboard.
There is no prize; no cash, medals, or Kaggle points are awarded. Entering the leaderboard is free, and no stake is required or accepted. The challenge focuses on whether a general-purpose language model can beat the parimutuel market, which aggregates millions of euros of bets into a single calibrated probability. Having the full Huggy archive makes the work easier but is not a condition to enter or to rank well.
Participants submit a ranking of five horses in order via a standard Kaggle CSV file. In each race, the five selected horses get ranks 1 through 5 in the predicted order, while every other runner gets rank 6. Submissions are scored on the Root Mean Squared Error (RMSE) between the predicted rank and the true rank. Lower is better, with the exact order scoring 0.00. Order matters as much as selection, mirroring the Quinté+ where order pays far more. Benchmarks include the Market favourite (PMU odds) measured on rehearsal-phase data.