So You Think You Can Predict the World Cup: Part 2
In the first part we saw the factors that Joachim Klement used to predict the outcome of the FIFA World Cup Tournament. Now, we see how these factors determine the 2022 winners.
Image courtesy: FIFA 2022 Finals
In the 2022 world cup, the initial groups were as follows:
The figures in the brackets are the probabilities of the team making it out of the group stage and into the round of 16. Since two teams will make it out, the probabilities add up to 200% instead of 100%. These probabilities have been calculated based on the factors discussed above.
This means Netherlands, Senegal, England, USA, Argentina, Mexico, France, Denmark, Spain, Germany, Belgium, Croatia, Brazil, Switzerland, Portugal and Uruguay would be the 16 countries to make it to the knockout stage.
A team’s likelihood of progressing to the knockout stage isn’t a guarantee that it will. Consider Japan in Group E, which has a 40% chance of moving on to the knockout round. This is less than Germany (55%) and Spain (67%) combined. Therefore, it is expected that Spain and Germany will advance through the group stage. However, if the World Cup were to be repeated repeatedly, Japan would typically get to the knockout stage four out of ten times (meaning either Spain or Germany won’t). It frequently happens that group favourites are ousted. It’s just how probabilities work.
How true were these predictions of the group stage? Turns out, only 5 out of 16 were wrong.
Mexico, Denmark, Germany, Belgium and Uruguay were replaced by brilliant performances from Poland, Australia, Japan, Morocco (the biggest surprise for the world, made history as the first African country to make it to semifinals) and South Korea.
Model’s predicted round 16 matches:
5 out of the 8 winners of the knockout round were predicted correctly. As mentioned earlier, Mexico and Germany were expected to but didn’t make it to the knockout stage and Morocco being the surprise factor in this tournament, defeated Spain, another country predicted to win.
Model’s predicted quarter finals:
This time only 1 out 4 of the model’s winners met the expectations. France beat England to the semi finals, Morocco beat Portugal and Spain did not even make it to quarters.
Model’s predicted semi finals:
Again, only 1 prediction turned out to be half true. Argentina won the semi final against Croatia.
Finally, the predicted final between Argentina and England actually happened between Argentina and France with Argentina winning as foretold.
The Argentine victory adds another feather to Klement’s hat. However, for those wondering why some of the most important predictions of the later stages of the tournament went wrong, here’s a plausible explanation:
This time, a few key aspects of a World Cup have changed. The World Cup is not hosted at the end of a lengthy season when the majority of stars have played more than 50 matches, which is crucial. Some players, like Robert Lewandowski of Poland, are more likely to be fit and performing at the top of their game than in prior World Cups. They are less worn out since they have played less games in their domestic leagues or the Champions League. They can therefore have a greater influence than usual. This again helps teams that largely rely on an individual player’s brilliance, like Lionel Messi or Christiano Ronaldo. The fact that these two men are no longer the youngest speaks much about them in particular. Christiano Ronaldo is 37, whereas Lionel Messi is 35. The gap between playing in a tournament after around 15 domestic games and playing in a tournament after about 50 domestic games is significant for these players.
The same factor works against certain teams. Normally, national teams have a period of several weeks to practise together before the tournament starts. This benefits teams that rely less on an individual star player and more on a well-oiled machinery of different world-class players harmonising with each other. Spain and Germany are such teams. This time, the World Cup in Qatar starts right in the middle of a season with little time for national teams to practise together and establish a routine.
There are also certain economic principles that explain failure of the model (Although Klement himself calls these “generic excuses”):
The ‘if only’ defence: Experts claim that they would have been correct if only their advice had been followed. For e.g. “If only France had made it out of the group stage, they would have made it into the last 16”.
Ceteris paribus: Means “other things equal”. Something outside of the model of analysis occurred; therefore, it failed. Say, Christiano Ronaldo got injured; therefore, the model’s prediction about Portugal failed.
I was almost right: Although the predicted outcome didn’t occur, it almost did. For example, Although England didn’t make it into the finals, they made it into the quarter-finals.
It just hasn’t happened, yet: Experts often claim “I wasn’t wrong, it just hasn’t occurred yet.” The model wasn’t wrong, it was just predicting another World Cup in Qatar.
The single prediction: “You can’t judge me by the performance of a single forecast.” You can’t judge the model by the performance of the 2022 prediction. Look at its predictions for 2014 and 2018.
There are other models that predict the outcome of the World Cup. Lloyd’s put together the collective insurable value of team players, based on the notion that salaries in elite football are a decent proxy for talent and they predicted England are going to win versus Brazil in the final. Although this model’s 2022 prediction went terribly wrong, they got 2014 and 2018 right.
Joachim Klement, when asked about this in a Bloomberg interview, said “I’m a bit cautious about using the collective value of the players because the premier league has so much more money at its disposal and can pay so much more for players that it is a bit skewed in that respect but there are different approaches I as an economist use the macro approach that has turned out to be rather better than I expected to be.”
This reason and the fact that no other model has predicted the final outcome correctly three times in a row is what makes Klement’s model the best of its kind as of today.