Skip to content

Improved Horse Race Betting using Regression and Classification Models

License

Notifications You must be signed in to change notification settings

juanjaho/horse-race-betting

This branch is 8 commits ahead of theianchia/horse-race-betting:main.

Folders and files

NameName
Last commit message
Last commit date

Latest commit

8558c7e · Apr 2, 2024

History

24 Commits
Apr 1, 2024
Apr 1, 2024
Apr 2, 2024
Apr 1, 2024
Apr 1, 2024
Apr 1, 2024
Apr 1, 2024
Apr 2, 2024
Mar 31, 2024
Apr 1, 2024
Apr 1, 2024
Mar 28, 2024
Apr 1, 2024
Mar 26, 2024
Apr 1, 2024
Apr 2, 2024

Repository files navigation

Investigating Alpha in Horse Racing

Project Report and Presentation

About the Project

This project aims to explore potential opportunities in the Hong Kong Horse Race Betting market, utilising data from Hong Kong Jockey Club.

Results

Best Performance of each Strategy

Mean Variance Strategy Machine Learning Strategy

  • Using a single-bet betting strategy where we only bet the horse with the highest probability of winning in a race, we are able to make substantial profits
  • Flat betting yielded the highest absolute profits
  • While using Kelly Criterion to perform bet allocation reduced absolute profits, average returns per bet consistently outperformed flat betting across all methods and target variables, except for Naive Mean Variance Strategy.
  • Overall, ML models performs the best. The ML models are able to predict the winner of a race more than 50% of the time

Overview of ML Model Profitability with Betting Strategy

Race Speed Regressors Finish Time Regressors Finishing Position Classifiers
Flat Betting $41,250 $31,290 $8,430
Betting with Kelly Criterion $30,360.08 $28,419.82 $4,914,48
Prediction Accuracy 57.3% 55.1% 53.9%

Using a Flat Betting Strategy

Using Kelly Criterion to allocate bet amounts

Model Training Performance

About

Improved Horse Race Betting using Regression and Classification Models

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 100.0%