Prediction of football scores using Gaussian Processes. We create independent models of home and away team scores using a sparse variational Gaussian Process (SVGP) with a fully-factorised Poisson likelihood and a Matérn 5/2 kernel. The bulk of the code is contained in the notebook model.ipynb with some helper functions in /src.

To build the environment, run the following commands in a shell:

$ python -m venv env
$ source env/bin/activate
$ (env) pip install -r requirements.txt

To create a Jupyter kernel, run:

$ (env) ipython kernel install --name "env" --user

Note: The setup instructions above have only been tested for python 3.10 on M1 Mac.

To run the jupyter notebook (model.ipynb), run:

$ (env) jupyter notebook