Sport Ratings

University of Amsterdam, Fall 2013
Research project proposal for Bachelor Artifical Intelligence (AI), Honours Programme

Supervisor: Christian Schaffner (ILLC / email)

Students: Joost Hoppenbrouwer & Marysia Winkels

Final Report & Code

Here is the final report and the code Joost & Marysia have produced in their project. During the project, we have been visiting Simon Gleave and his research team at Infostradasports in Nieuwegein. Sounds interesting? Check out the follow-up project!

Motivating Problem

In American Football, the 120 top US-college teams each play a season of roughly 15 games against some geographically close opponents. Given the outcomes of all these games, determine the top two teams that deserve to play in the national college super bowl.

The standard way of ranking teams based first on number of wins and then on goal difference might work rather poorly in this case, because not a full Round Robin has been played. Also, this system does not take into account the strengths of the opponents. Sport rating systems analyze the results of sports competitions to provide objective ratings for each team.

Research Objectives:

Understand the mathematics and rationale behind different sport rating systems such as power rankings (used in Ultimate Frisbee) or the ELO rating system (used in Chess).
Figure out the advantages and disadvantages of different rating systems.
Visualize the rating systems with meaningful graphics such as this one.

Approach:

Read and understand the relevant mathematical literature such as this thesis.
Write a report describing the investigated systems, listing their advantages and disadvantages.
Use graphical libraries such as D3.js in order to come up with instructive interactive graphs to illustrate how the rating has been computed.

Presentation