New Kaggle Competition: World Cup 2010

Can you outdo the quants?

As a break from projecting the strength of subprime mortgages, credit default swaps and other abstruse financial instruments, quantitative analysts at Goldman Sachs, JP Morgan, UBS and Danske Bank have modeled the 2010 soccer FIFA World Cup [see below - Ed]. We have set up a forecasting competition, allowing competitors to go head-to-head with these corporate giants.

The challenge is to correctly predict how far each country will progress in the tournament. There is a small dataset and some links to variables of interest on the data page. There are links to the investment banks' predictions and other World Cup modeling efforts on the hints page.

We are running two challenges side-by-side - a Take on the Quants Challenge and a Confidence Challenge. The Take on the Quants Challenge simply requires competitors to pick how far teams will progress in the tournament. Competitors' entries will be ranked against the predictions made by the investment banks. The Confidence Challenge requires competitors to assign a level of confidence to each prediction - a competitor's score is weighted by their level of confidence. Competitors enter both challenges with a single submission. There are more details on the submission instructions page and the evaluation page.

The competition closes just before the first game kicks off on June 11th.

Compete at kaggle.com/worldcup2010

This blog posts gives more information about the bankers methods

UBS and Danske Bank use econometric models to find variables that explain World Cup success. UBS find recent record to be important. On this score Spain looks promising, having won every single qualifying game. Also important is where the cup is hosted - a third of all tournaments have been won by the host nations and those playing on their own continent tend to perform well.

Danske Bank modelling accounts for socio-economic variables such as population size and GDP (bad luck Greece). According to Danske, the wealthier countries tend to perform better as do those with larger populations - but this effect diminishes with population size. However, the Danske model has some odd features - for example they find FIFA ranking correlates negatively with a country's performance.

Goldmans and JPMorgan's approach is somewhat different. Rather than trying to find the characteristics of a World Cup winner, they look for variables that correlate with World Cup success. They take into account betting market odds and FIFA ranking. JP Morgan also compare odds across betting markets as a proxy for the certainty of punters' convictions about different teams.

Goldman Sachs, UBS and Danske Bank are all predicting Brazil to win, while the quants at JP Morgan tip England to hoist the trophy on July 11.