A personal project called Hoofs has been developed to tackle horse racing as a machine learning ranking problem, focusing on British and Irish races. The project has compiled approximately 1.18 million historical runner records and a feature bank of around 1,700 potential signals per runner. Models estimate win and place probabilities, with evaluation conducted chronologically using walk-forward validation to prevent future information leakage. While the model-only win AUC is around 0.729, the market baseline shows a stronger AUC of approximately 0.790, highlighting the challenge of extracting information not already reflected in market prices. AI
IMPACT Demonstrates the application of ML ranking and walk-forward validation to complex, real-world prediction problems.
RANK_REASON The item describes a personal project applying ML techniques to a specific domain (horse racing) and presents results, fitting the research category. [lever_c_demoted from research: ic=1 ai=0.7]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
