[Currently a work in progress]
This project aims to provide weekly predictions for the English Premier League through a simple API, accompanied by an interactive Dash Dashboard.
The project utilizes publicly available data from various sources to produce odds for Home/Draw/Away for upcoming games. These odds can be used to spot inefficiencies in betting markets, which provide the potential of profiting from them.
An XGBoost model is used, built inside a modular, re-usable project architecture that I have designed and use in my own projects. Feel free to copy what you find useful, and if you ahve any questions please dont hesitate to contact me at qemtek@gmail.com
The main features of the project are as follows:
- Automated CI/CD workflow - Built with CircleCI. Every time an update is published to the master branch, tests are automatically ran and the new package version is built/uploaded to Gemfury. Each day the model is scheduled to make predictions and each week the model retrains, saving the best performing model.
- API to serve predictions - The API loads the model package (not the source code) to serve predictions. This isolates the behaviour of the model from the behaviour of the API, ensuring we do not make any predictions with development code.
- Model monitoring dashboard - Built with Dash. This dashboard shows the cumulative profit, versus betting on the favourite, accuracy grouped by different teams/scenarios and more. This dashboard is hosted on a Heroku server (link coming soon).
Thanks for visiting!
Environment variables required for the project to run:
Project Specific Credentials
- PROJECTSPATH - The path to the project (root/packages/football_trading)
- DB_DIR - The location of the DB locally (set to {your_project_root}/db.sqlite)
- RECREATE_DB - Whether to do a full refresh of the database
- LOCAL - Whether to store data locally or in S3
- IN_PRODUCTION - Whether the model is being run in production (set to false normally)
- PRODUCTION_MODEL_NAME - A prefix used to select the right model if multiple models are in production
S3 Credentials
- S3_BUCKET_NAME - The name of the S3 bucket you want to use for storing data
- AWS_ACCESS_KEY_ID - Your AWS access key id
- AWS_SECRET_ACCESS_KEY - Your AWS secret access key
Betfair Exchange API Credentials
- BFEX_USER - Your betfair username
- BFEX_PASSWORD - Your Betfair password
- BFEX_APP_KEY - Your Betfair Exchange API app key (you need to request one from Betfair)
- BFEX_CERTS_PATH - Path to a folder containing your Betfair Exchange API certs. You can find a limited set that I own in the root directory under /certs.