On Wednesday 27 June, the German football team lost to South Korea 2 – 0 in the 2018 World Cup. As a result Germany finished bottom of their qualifying group, and so will not progress to the knock-out stages. It was an unexpected result as Germany won the last world cup in 2014 and are ranked number 1 in the world by FIFA. South Korea´s best result is final 4 in 2002 (held in Japan and Korea) and are currently ranked 57.
Germany´s departure from the World Cup before the knock-out stages caused headlines around the world, especially in Germany. Bild´s headline was “No Words” to express their shock, and the Berliner Kurier went with “We mourn for shame”. So is this event the biggest upset since Trump´s election and a national tragedy for Germany or just a minor underperformance by a pretty good football team?
I´ve talked before in this blog about the difficulty in predicting the future in “Prepare to be wrong“. In terms of the research it´s also worth mentioning that, in general, we tend to overestimate our ability to make predictions. Philip E. Tetlock a researcher and author published this finding in his 2005 book: “Expert Political Judgement: How good is it? How can we know?”. By asking experts and non-experts to make predictions and then comparing the results to those predictions made by pure extrapolation he found that extrapolation provided the most accurate predictions (still not that accurate though). Non-experts were the least accurate, and experts were somewhere in-between. In other words none of these groups or methods were able to provide consistently accurate predictions , and experts were only slightly better than random “guessing” or non-experts. There was one area where he found experts were an order of magnitude better than non-experts. As well as asking his respondents to make predictions he also asked them to estimate the probability of those predictions being correct. In general, experts put far lower probabilities on the accuracy of their predictions than non-experts. So although experts were only slightly more accurate in their predictions, they were much more realistic in understanding the likelihood of their prediction becoming reality.
What does this tell about Germany and the World Cup? Well, we know that Germany have won the World Cup 4 times since it´s inception in 1930. We know the tournament has been run 21 times. We also know that before the tournament the bookies put them as second favourites at 9/2. This tells you that experts believe it is more likely that Germany will lose the World Cup than win it – predicting that they should at least reach the quarter finals. History would suggest the same. It´s worth mentioning that “favourite” in betting parlance is a bit of a misnomer. In betting terms, most of the time, “favourite” means “will probably lose but is less likely to lose than the other contenders”. Much has been made in the press of this being the first time Germany have failed to make it out of the group stage, but this is probably less significant as the tournament has changed format significantly over the years with more teams being added. 32 teams were included in the World Cup for the first time in 1998 so the sample size is much smaller and it´s harder to draw conclusions about the significance of this result. So, statistically, Germany underperformed but, in terms of World Cup history, perhaps this isn´t as significant as the headlines would have us believe.
When Tetlock did his research he looked for patterns in his data. Was there any particular demographic (age, gender, geography, education etc…) that produced better predictions? He searched in vain – all were equally poor. He did, however, find one factor. Experts with a higher public profile (wrote in popular newspapers, regularly appeared on television) tended to have worse accuracy in predictions than others.
The psychological explanation of this lies in the power of stories. Humans like drama: true stories that we connect to emotionally. A dramatic or shocking prediction is likely to attract a larger audience and therefore build a higher profile for the forecaster than a dull but more accurate prediction. This also explains our reaction to Germany. We remember the high drama of the 7-1 defeat of Brazil in 2014, and the penalty shoot-outs won by Germany in the past. We give them more weight in our memory and so a couple of “out of sorts” results against Korea and Mexico seem all the more shocking. Now, we start again. The papers are already writing the “Where it all went wrong” stories recasting the history of Germany at this World Cup to make their defeat seem inevitable, whereas two weeks ago the same papers were confidently predicting a serene cruise to the final. Perhaps a more balanced view is to see this as a “reversion to the mean”. After all, Germany have lost the World Cup many more times than they have won it. I´m sure they will be back.