Thursday 22 July 2010

How to win at World Cup predictions

Now the World Cup is over, the sound of the Vuvuzelas is fading almost to a melodic memory, the England shirts given away to the local charity shop, it's time to round up the success or otherwise of the various World Cup predictions, based on mathematical techniques.

Firstly, let's look at the banks. I have no idea why our esteemed financial institutions have employed people to make football predictions, when they should really be working rather hard at recouping the money we've given them. But here goes:

UBS failed dismally with a prediction that Brazil would win, and another that Italy would get quite a long way.

Goldman Sachs also backed Brazil, predicting the other three semi-final teams to be England, Argentina and Spain. One out of four there, and no winner.

Danske Bank, not wishing to disagree, backed Brazil too, but JP Morgan must be feeling really silly. They tipped England for the title.

And a quick look at some of the academics. Metin Tolan, a German physics professor, thought that Germany would end up victorious:


The most successful analysis we've seen so far has been that of the mathematics department at Queen Mary, University of London. They analysed each game's pattern of passing, giving each player a score called their 'centrality'. This measures how important each player is to the team's performance. Similar techniques are used in modelling computer networks.


From this analysis, the researchers say they predicted a Spanish win in the final against the Netherlands, although it is not clear from the website how exactly they reached this conclusion. Also, it seems impossible to predict the winner of a tournament before knowing which teams would play against each other in the latter stages.

What does all this tell us? The world is a very unpredictable place and you'd probably be better off becoming a bookmaker than visiting one. That or buy a psychic octopus.