We’ll go ahead and say it: it’s been quite a peculiar World Cup.
Brazil’s humiliation at the hands of Germany in Belo Horizonte will go down in fútbol annals. The score of 7-1 was the heaviest defeat suffered in the semi-final of a World Cup, ever – and was the game that defied all prediction analysis models.
What started out as a pulsating affair turned into a funeral setting as the Germans scored five goals inside the first half hour. Unfortunately for Brazil, the second half wasn’t much different.
As if the game itself wasn’t shocking enough, the results left data scientists and thoe touting the successes of prediction analyses scratching their heads: Nate Silver included.
The outcome of sporting events, however, elude even Silver, proving that perhaps sports are wholly unpredictable.
Silver, the celebrated statistician behind the FiveThirtyEight blog, had pegged Brazil as the favorites to win the cup and slated their chances of success against Germany at 65%.
Silver has an infamous history of using data to predict the outcomes of seemingly unpredictable events. His name shot to fame when in 2012, he perfectly predicted the presidential election winners of all 50 states.
The outcome of sporting events, however, elude even Silver, proving that perhaps sports, in which an individual or team competes against one another to outshine the other’s physical exertion and skill, are – dare I say it – wholly unpredictable. Everything can change, if human ingenuity and hard work have anything to do with it. Yes, even data scientists cannot predict human anomaly, and sports illustrates this best.
Neither Data Scientist Nor Bank…
This prediction-upsetting phenomenon has become somewhat of a pattern at this year’s World Cup. As you may remember, Goldman Sach’s economists crunched data on 14,000 past matches to arrive at a 67-page report that confidently calculated the outcomes of all matches, published May 30.
The bank changed 50% of the teams after the first stage once it became clear that reality wasn’t quite matching the historical dataset.
Looking at those original, pre-tournament predictions, only 37.6% of the group stage matches ended the way Goldman Sachs forecast. That is, it correctly calculated one of three possible outcomes — win, lose or draw — just over a third of the time, hardly better than predicting it at random.
Even data scientists cannot predict human anomaly, and sports illustrates this best.
When the bank updated its predictions at the end of the group stages, it got the result of the match — win, lose or draw — right 60% of the time. But, it failed to correctly guess the results of both semi-finals at that stage, and hence ended up predicting the wrong teams for the final.
So it seems neither data scientist nor bank can properly predict athletics.
Here’s what they found:
- Of the predictors analyzed, all bet on Spain and missed
- Nobody bet on Costa Rica and they made it to the round of 16
- An accounting firm (PwC) out-predicted a sports network (ESPN)
- Goldman Sach’s acclaimed prediction came in last
Who has the best predictions thus far? Bloomberg, with 62% correct positions and 68% correct countries. Even then, Bloomberg’s success rate has remained less than perfect.
So there you have it: data science is anything but a concrete science. Yes, even we admit it.
If the Brazil v. Germany game taught us anything, it’s that Sunday’s final is anyone’s game.
Any way you slice it, the World Cup final between Germany and Argentina should be superb. The story lines are certainly plentiful: A European team trying to win the World Cup for the first team on South American soil, and an Argentinian team trying to win a World Cup on the soil of their bitter rivals, Brazil. Arguably the best player in the world, Lionel Messi, taking on a German squad that has played like a well-oiled machine. Two countries that are competing for the third time in a World Cup final (Argentina won in 1986, West Germany in 1990).
It’s going to be a heck of a match. If you’re still checking the predictions, Nate Silver’s Soccer Power Index (SPI) match-predictor, has Germany favored to win by 63%. And Bloomberg? They predict Germany wins by a mere 50.9%. Of course, if the Brazil v. Germany game taught us anything, it’s that Sunday’s final is anyone’s game.
Best of luck to both teams and happy watching!