Soccerstan

Reproductions of models for football (soccer) matches in Stan/PyStan.

How do I run the models?

It's pretty simple. You can just run src/soccerstan.py from the command line. This requires the file containing match data (from football-data.co.uk) and which model you want to use as arguments.

For instance:

> python soccerstan.py 'data/example.csv' 'karlis-ntzoufras'

Models

So far, the following models have been implemented:

• Maher (1982) - Modelling Association Football Scores - maher
• Dixon and Coles (1997) - Modelling Association Football Scores and Inefficiencies in the Football Betting Market - dixon-coles
• Karlis and Ntzoufras (2008) - Bayesian modelling of football outcomes (using the Skellam's distribution) - karlis-ntzoufras

Data

The main soccerstan script has been written to work with data from football-data.co.uk. The 2015/16 English Premier League season has been added as an example data file.

However, it will also work with any csv with columns labelled home_team, away_team, home_goals, away_goals.