Machine learning the hard way -- a story about ponies
summary
Buys $800 of DRF race data, builds an ugly parser, mungees into numpy arrays, and uses scikit-learn with grid search plus a custom betting scorer. Discovers the raw odds feature dominates models, standardizes features, switches from classification to regression and uses SGD regressor to predict true win probabilities. Simulation shows a small edge (single‑digit ROI overcoming the 20% take) but real betting fails at scale because odds move, pools shift and wagers must stay tiny to stay profitable.
AI-written summary · about
chapters (13)
- 0:00Introduction to ponies
- 1:36How horse racing works
- 4:02Machine learning basics
- 7:39Data management challenges
- 12:54Data munching with numpy
- 15:15Using scikit-learn
- 18:34Standardizing data
- 20:24Refining the objective
- 21:54Selecting algorithms
- 26:51Automating pipelines
- 30:27Validating model results
- 33:37Lessons and reflections
- 37:09Q&A session
related talks
32mHow I lost 1000€ betting on CS:GO with machine learning and Python
23mWhy Fraud Detection, Sports Betting, and Bitcoin Prediction Look Alike
44mWinning with Simple, even Linear, Models
53mDeep Learning in the world of little ponies
33mYour best Bet: Effortless MLOps with Python Models in
35mIntelligent Ruby: Getting Started with Machine Learning
01h 02mDeep Learning in the world of little ponies
30mMachine Learning: Beyond the Tutorial