Today the FIFA World Cup begins. But: Who will win the World Cup? This question plagues millions of fans worldwide. Among them also scientists. Therefore, four statisticians from Germany and Belgium have developed a model and calculated with what probability each team could win the title. The result: The German football national team will become World Champions with a probability of 17.1%. Top favorite for the title is Spain (17.8%), while Brazil (12.3%) has rank three and is followed by France (11.2%).
Starting the round of the last sixteen, Germany is favorite
Interestingly enough however, Germany is actually the strongest team according to the model. In a direct duel with Spain or Brazil, the team of Joachim Löw would have the higher probability of being the winner. Germany, however, will first have to play against opponents who tend to be stronger than those playing against Spain. This becomes particularly clear in the round of the last sixteen, where Spain plays against a team from group of A (Uruguay, Russia, Saudi Arabia, Egypt) and Germany plays against a team from group of E (Brazil, Switzerland, Costa Rica, Serbia). While Spain and Germany have similar probabilities of reaching at least the round of the last sixteen (88.4% or 86.5%), the differences in probabilities for reaching at least the quarterfinals (73.1% or 58.0%) are quite clear. If Germany were to reach the quarterfinals, the team, starting at this time, would be the favorite to win the title.
Simulation of 100,000 tournament processes
The researcher team consisting of Dr. Gunther Schauberger (Chair of Epidemiology of Prof. Dr. Stefanie J. Klug), Prof. Dr. Andreas Groll (Technical University of Dortmund), Prof. Dr. Christophe Ley and Hans Van Eetvelde (both of Ghent University) produced 100,000 simulated tournament processes, from which the probabilities were calculated. "In each individual simulation, each of the 48 matches of the group stage was prognosticated in accordance with our model and, building on those results, the further tournament process up to the final World Champion is predicted," Schauberger explains. A part of the researcher team had provided and already prognosticated a forecast for the 2014 tournament, with Germany predicted as being World Champions.
The model used is a combination of a so-called random forest (a technology from machine learning) as well as a ranking approach, for which the current strength of each team was computed similarly to that in the FIFA ranking list. "Concrete aspects which go into the modelling are, for example, the quotes for each team among the bookmakers, the current rank in the FIFA world rankings, the success of the players of each team in the preseason - the latter is operationalized through the success in the Champions League," describes the doctor of statistics.
During the 2014 tournament, it turned out that the forecast of the group with regard to all of the WM games was more precise than the prognoses from the bookies. "Mathematically, the first predicted calculation of the WM only differs slightly from epidemiological methods that are used to predict epidemics, for example a flu wave," says Full Professor Klug. The calculation of the WM is naturally, in a certain sense, a mere plaything, although it nevertheless also shows the breadth of the subject and that which is possible with solid statistical methods, says Prof. Klug and adds, "I naturally hope that Germany will become World Champion."
To the Homepage of the Chair of Epidemiology
Complete publication "Prediction of the FIFA World Cup 2018 - A random forest approach with an emphasis on estimated team ability parameters"
Contact:
Dr. Gunther Schauberger
Chair of Epidemiology
Georg-Brauchle Ring 58
80992 Munich
Email: Gunther.Schauberger(at)tum.de
Text: Fabian Kautz
Photo: Pixabay; Groll, Ley, Schauberger & Eetvelde