This Machine Learning Model Picked Spain to Win the 2018 World Cup

Statisticians at German technical university Technische Universitat Dortmund built a model that used machine learning to predict Spain will win the 2018 World Cup. The prediction is based on 100,000 simulations of the tournament. Spain was followed by Germany, Brazil, France and Belgium in terms of their chances of winning. And it should be a good tournament because Spain, with a 17.8 percent chance of winning, is only slightly ahead of Germany at 17.1 percent. Brazil follows with 12.3 percent, and then it's France (11.2 percent) and Belgium (10.4 percent). The researchers at the (German) university also said the model showed Germany is the strongest team and would win in direct competition with Spain or Brazil, but it has a harder road to the final. They pointed to the round of 16 in particular, noting Spain will face a Group A team (Uruguay, Russia, Saudi Arabia or Egypt), while Germany will face a most likely tougher Group E team (Brazil, Switzerland, Costa Rica or Serbia). TU Dortmund published the findings in a new paper that also determined Uruguay and Russia will top Group A; Spain and Portugal will lead Group B; France and Denmark will rank highest in Group C; Argentina and Croatia will win Group D; Brazil and Switzerland will dominate Group E; Germany and Sweden will come out on top in Group F; Belgium and England will triumph in Group G; and Colombia and Poland will lead the way in Group H. And, unfortunately for fans of Saudi Arabia, the researchers found it has a 0 percent chance of reaching the final. The research included three models: a regression approach previously used for the 2012 and 2016 UEFA European Championships and the 2014 World Cup; a random forest, which TU Dortmund said is a machine learning technique; and a ranking approach similar to FIFA's rankings but based on a statistical model. The first two models are based on variables like the teams' odds of winning the World Cup, FIFA rankings, average player ages and the success of players in the UEFA Champions League. The research looked at each team's probability of winning as well as surviving each round and the most likely results for the overall tournament. Researchers also used data from the four previous World Cups to calibrate the models. And, they said, the random forest and ranking approach came very close to the predictive quality of bookmakers. To further improve the prognosis, researchers added team skills. TU Dortmund said this increased the predictive quality significantly, and the statisticians were able to beat the bookmakers in the games of the last four World Cups. And to “further validate the predictive power of the model,” TU Dortmund said, researchers will bet “a small amount” of money on all 64 games. TU Dortmund is not alone in applying AI to the 2018 World Cup or using whatever methods they can think of to predict the winner. Goldman Sachs, for example, used AI to analyze the number of goals scored in each match, according to The New York Times, but it determined Brazil will win. And then, of course, there are more traditional methods. In 2010, Paul the Octopus, a mollusk at the Sea Life Centre in Oberhausen, Germany, gained notoriety for his accurate World Cup predictions, but he died a few months later. In 2018, Russia is reportedly hoping a deaf white cat named Achilles will follow in Paul’s tentacle-steps. The Psychic Pets website is also giving people a chance to sign their pets up to make predictions. Participants include Boe Boe the chicken, who supports Belgium; Adele the tortoise, who supports Switzerland; Lil Johnny the rat, who supports Denmark; and Sitka the ferret who supports Germany. Those who don’t have pets—or who don't trust AI, traditional bookmakers or their own intuition—may also want to consider time travel.