Spring 2012 Seminar Series

DATE: March 1st
LOCATION: E51-335
TIME: 4:15pm
Reception immediately following in the ORC Conference Room (E40-106)

SPEAKER:
Nils Rudi

TITLE
In-play football (soccer) prediction

ABSTRACT
For in-play football prediction, we consider a model based on Poisson arrival rates of goals. We prove that for a given pre-game trinomial predictive distribution of the result (i.e., home, tie and away), there exist a unique pair of Poisson arrival rates which correspond to it. This facilities the base model of in-play prediction with an absolute minimum requirement of data analysis. This result is extended to non-stationary arrival rates and state (score difference) dependent arrival rates. We perform an empirical investigation of the models using a large set of data from multiple leagues, and contrast the results with those of logistic regression.

 

In the presentation, I will also talk about multiple research opportunities in football, and current work we are doing with an extensive data set from the English Premier League.