How to Predict the Correct Result in over 50% of World Cup Games

In my last blog I made some quite strong criticisms of the way that FIFA calculate their rankings for international football. That was my opinion, but I didn't back any of it up with facts. I decided to do a sort of experiment to see how effective the FIFA rankings are in predicting results in a World Cup. Obviously I don't know the results for this year's competition so I have looked at the results in the last competition. I took the ranking for each country that played in the 2010 World Cup at the start of the competition and for each game I compared these ratings for each side. So for example in the first game when hosts South Africa played Mexico, I made the “prediction” for an away (Mexico) win, because Mexico were ranked 24th in the world compared to South Africa's 66th. On this occasion the game was actually a draw, so the FIFA prediction was wrong. In the next game, France (ranked 21st) played Uruguay (ranked 6th), this game was again a draw after the FIFA rankings had predicted a Uruguay win. In fact the first three games all ended with different results to how the rankings predicted (two of them were draws, but when the system is changed to give a draw where rankings are close, this actually makes it less successful). As can be seen from the following table, the FIFA ranking predictions then started to improve.

I don't have the odds that were available on games in the World Cup 2010, but my gut feeling would be that it would be borderline whether or not any money would have been made using the system I have described during the 2010 competition. Some of the games would have been between even money and 2/1, but I think there would have been balanced out by at least as many games that were below even money. North Korea, Honduras and Algeria are just three teams who I am certain would have faced opposition shorter than the 1.5 mark to win in each of their three games (although of course Switzerland and England respectively were kind enough to offer Honduras and Algeria a point in one each of their games).

When I started to compile these statistics I didn't know what the results would be. To be honest I probably expected to FIFA rankings to be less reliable than they turned out to be. However, it is difficult to say how well they have done really given that I don't have the odds that were available for the games, and as yet I haven't offered an alternative system, to see whether that is better of worse. Should anyone be in possession of a dataset for odds on these games, I would be very interested to see it. Otherwise I plan to suggest an alternative ranking system to the FIFA one and see if that can do a better job of predicting the results.

This is the third in a series of blogs about international football rankings, you can read the previous blog here.