© Pixabay
Author profile picture

It doesn’t look good at all for the German national team. According to statistical models developed by scientists at several European universities, Jogi Löw and his boys will probably go away empty-handed at the European Championship, which starts on Friday, June 11, and be able to pack their bags early. An international team of researchers consisting of Andreas Groll and Franziska Popp (both from the Dortmund University of Technology, Germany), Gunther Schauberger (Munich University of Technology, Germany), Christophe Ley and Hans Van Eetvelde (both from Ghent University, Belgium), Achim Zeileis (University of Innsbruck, Austria) and Lars Hvattum (Molde University of Applied Sciences and Arts, Norway) has used machine learning to calculate that France will probably win the 2018 European Championship title after winning the World Cup.

Using statistical models with information about team structure (e.g., market value, number of Champions League players, club match performance of individual players) and socio-economic factors of the country of origin (population and gross domestic product), they simulated the entire European Championship 100,000 times match by match. They followed the tournament draw and all UEFA rules. Based on all the probabilities of all teams advancing to the individual rounds of the tournament, France was ultimately the first contender for victory with a win probability of 14.8 percent. England followed in second and third place with 13.5 percent and Spain with 12.3 percent.

Forecasts can also be off the mark

The researchers had to admit, however, that this outcome of the tournament is anything but certain. After all, the margins at the top are decidedly close – and the probability of winning, at just over ten percent, is quite low anyway. “It’s in the nature of forecasts that they can also be off the mark – otherwise soccer tournaments would also be very boring,” says Achim Zeileis. “We just provide probabilities, not certainties, and a 15 percent probability of winning means at the same time that the team cannot win the tournament 85 percent of the time.”

Nevertheless, past results show that the forecasts have been quite accurate. Back in 2008 at the EURO final and in 2010 and 2012 at the World Cup and European Championship, Achim Zeileis’ Innsbruck model, which is based on adjusted odds from betting providers, predicted Spain as world and European champions. For this year’s European Championship, the scientists are using the model as part of a more comprehensive combined model. This was developed by the teams led by Andreas Groll ( Dortmund University of Technology), Gunther Schauberger (Munich University of Technology) and Christophe Ley (Ghent University) and proved to be more accurate in its predictions than the betting providers at the 2018 World Cup.

And Germany?

After the draw, the German national team led by captain Manuel Neuer already faces some challenges in the group matches. “There are three very strong teams in Group F, including reigning world champions France and European champions Portugal, both of whom were also EURO 2016 finalists, plus Germany,” explains Andreas Groll. “In this group, the probability of making it to the round of 16 is therefore lower compared to the favorites in the other groups. But whoever manages to do that then has a pretty good chance of advancing.”

According to the forecast, both Germany and Portugal have an 85.3 percent chance of making it at least to the round of 16. For France, by the way, the odds are 89.7 percent according to the model. And the chances of Germany winning the title? A meager 10.1 percent. On par with Portugal.

An easier draw for Austria, but…

Austria’s national team should get through the preliminary round quite handily in the group matches against the Netherlands, North Macedonia and Ukraine. This stands in contrast to 2008 and 2016. “The favorites in the group are clearly the Netherlands, even according to our model, but after that comes Austria, which already has a probability of 80.9 percent of reaching the round of 16,” explains Achim Zeileis. “That is significantly more likely than for Ukraine and North Macedonia.” However, the team is unlikely to get much further, according to the forecast. The chances of bringing the trophy to Austria are just 1.5 percent.

Machine learning

The model is based on four sources of information: From Ghent University comes a statistical model for the playing strength of each team based on all international matches over the past eight years. The University of Innsbruck contributed a statistical model for the teams’ playing strength based on the betting odds of 19 international bookmakers. Finally, from the Dortmund University of Technology and the Munich University of Innsbruck came further information about the teams, for example, their market value, their countries of origin and population figures. Also, from Molde University of Applied Sciences, information was included such as detailed ratings of each player and their individual performance, both in their parent clubs and national teams.

A machine learning model served as the fifth source or “partner,” combining the other four sources and optimizing them step by step. To do this, the researchers previously trained the model with historical data. “We fed the model with the current data for the past four European Championships, i.e., between 2004 and 2016, and compared it with the actual outcomes of all matches in the respective tournaments – this way, the weighting of the individual sources of information for the current tournament will ideally be very accurate,” explains Andreas Groll.

The answer to the question of how accurate the model’s predictions are will be revealed at the latest at the final on July 11.

The entire forecast with interactive graphics is available here: http://bit.ly/forecas

More articles on machine learning can be found here.