How to predict a football match (slides)

stats and snakeoil

2018-10-13

Dixon-Coles, Expected goals, neural networks, regista, slides, xG

I was very fortunate to be asked by Dan Barnett if I was interested in doing a talk at his Improving with Data event.

I presented a talk on the Dixon-Coles model (pretentiously titled How to predict a football match). I thought I’d post the slides here for posterity.

Some of it may not be that clear without the commentary; although most of the insight is available in the original Dixon-Coles paper, or in content that is (or will be soon) posted on this blog (1, 2, 3).

Event description:

How do you predict a football match? Exactly how much better are Manchester City than Manchester United? How likely are Arsenal to beat Liverpool? We can begin to answer these questions with statistics. In this talk, Ben will walk through a foundational model in football analytics: the Dixon and Coles model and will explain how it works, how it can be used, and some different ways it can be developed for the modern era of xG and Big Data. Ben is a data scientist based in London. After 2 years working in sports betting an analytics, he now works at Wefarm, the world’s largest network for small-scale farmers.