The betfairValue application and website
How the site was created
The website was built using Python's Django framework. Django combines the functional power of Python to connect a PostgreSQL database with HTML/CSS front end.
Market Odds are extracted from Betfair Exchange using the Betfair API. Scheduled regular requests ensure odds presented on this site are near real-time. Processing is automated using the Python Apscheduler library and the data is stored in the PostgreSQL database.
Historical results and statistics are captured each night to ensure team strength statics reflect all completed matches. Python Scipy, XGBoost, Keras and Tensorflow libraries are used to generate a range of models each with their own set of result probabilities. The final probabilities published on the site is an ensemble of models which have demonstrated profitability in backtesting. Some of these models take the market odds extracted from Betfair Exchange as an input. Therefore, probabilities are recalculated after each market odds extract and may vary slighltly over time. As markets become more liquid and tighter closer to match kick off time, probabilities before the match start time should be considered the best model probability generated.
The application was deployed to a Linux server using Salesforce's Heroku cloud platform with physical storage in the Amazon EU West 1 data centre.