Modelling and predicting football
I’ve personally done a pretty poor job so far of predicting the results of the 2010 football World Cup. Perhaps the “quants” at JP Morgan and Goldman Sachs have done better? Kaggle is running a prediction contest, Take on the Quants, where anyone could pit their predictions against those of the quants. The competition is closed but the results aren’t out yet.
At the same time, Norwegian Computing Center is publishing a daily list of simulated probabilities of each team winning the World Cup. They use a simple statistical model, the parameters of which are updated in response to actual results. Currently Netherlands is the narrow favourite to win, with an estimated probability of 18.1%.
For those who are *really* into betting on sports, Win The Trophy has launched a sports betting API with which programmers can create “BetBots”, automated betting strategies that can be tested without using up actual money. Betting strategies, of course, need data, and for football those could perhaps be provided in part by Footbalistic, a football data site that aims to be a leader in football statistics and analytics. In a VentureBeat article, one of the founders describes Footbalistic as the first data-driven football site and explain how it has aggregated data from various sources into a queriable form.
The most fun football-related analysis of the last few months, though, is a PLoS ONE article by Amaral and co-workers. In Quantifying the Performance of Individual Players in a Team Activity, the researchers have managed to come up with a way to describe a football match, including the actions of individual players, in a graph. This was done using detailed data on passes between players, shots and so on; apparently, an unprecedented amount of this kind of data was collected during the European Cup of 2008. The approach can be used to quantify how well a given player performed in a game. In a ScienceNow feature, professor Amaral (who is Portuguese) quips that he will use the graph approach to analyze the disappointing (according to him) game between Portugal and the Ivory Coast, which was tied 0-0 – “maybe it turns out we played better than I thought.”