Skip to content

This repository explores non-standard models for football predictions, including Elo, TrueSkill and Neural Networks that were developed by me as part of the final year project at the University of Surrey.

License

Notifications You must be signed in to change notification settings

Gw1p/fb-prediction

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Using Non-Standard Models for Football Predictions: Elo, TrueSkill, Neural Networks

by Anton Bendrikov

Supervisor: Prof Paul Krause

All the code is within a single Jupyter Notebook ab01719.ipynb.

The file should be followed from the top to bottom. It starts by importing all the data, running tests and pre-processing.

Then, the file is split in sections with appropriate headings:

  • Elo-related operations can be found under Elo
  • TrueSkill model is developed under TrueSkill
  • Neural Network is the largest section and contains all the workings for developing an MLP

Models' sections generally parse the data into correct format and setup the required dependencies, then grid-search the relevant parameters before presenting final evaluation.

Set Up

Data

In this project, data for England Premiership, France Championnat, Germany Bundesliga and Italy Serie A is used. Additionally, NBL, NFL, Super Rugby and Twenty20 Big Bash data is utilised when evaluating the Neural Network.

The relevant data can be downloaded from:

Formatting the Data

Once the relevant data is downloaded, it should be placed in the relevant folder under source/Data/<competition_name>.

The file names need to be formatted as follows:

  • England Premiership: Permier20XX-20YY.csv
  • France Championnat: Championnat20XX-20XX.csv
  • Germany Bundesliga: Bundesliga20XX-20XX.csv
  • Italy Serie A: SerieA20XX-20XX.csv
  • NBL: NBL20XX-20XX.csv
  • NFL: NFL20XX-20XX.csv
  • Twenty20 Big Bash: TwentyBigBash20XX-20XX.csv
  • Super Rugby: SuperRugby20XX-20XX.csv

Alternatively, modify calls to get_sport_data to format file names according to a different format.

Dependencies

See requirements.txt

About

This repository explores non-standard models for football predictions, including Elo, TrueSkill and Neural Networks that were developed by me as part of the final year project at the University of Surrey.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published