According to the analysis conducted by the researchers, France, Croatia, Denmark and Hungary will win the four preliminary-round groups. There is a 41.2 percent chance that Denmark will become world champion; the probability of a victory by France or Croatia is 18.4 or 9 percent, respectively. The model puts the chances of an overall German victory at 6.4 percent.
Data analysis has become popular in sports, among other things for scouting and in determining betting rates. Bookmakers currently see primarily Denmark and France as the favorites for the handball title – as does the scientific model. The team led by Prof. Dr. Andreas Groll of the TU Dortmund Chair for Data Analysis and Statistical Algorithms joined Dortmund graduate Jonas Heiner, Bochum handball expert Jörn Uhrmeister and Dr. Gunther Schauberger from the TUM Chair of Epidemiology.
A Prediction Algorithm from epidemiology
The algorithm used by the scientists is similar to the procedure used to predict the course of epidemics. After the announcement of the provisional group of 28 for the handball world championship preliminary round, the team simulated the outcome of the tournament including hundreds of thousands of different combinations of game results; they calculated a probability for each possible game outcome. “In every individual simulation, each game of the preliminary group phase was predicted in accordance with the model,” explains Gunther Schauberger. “Based on this, we then predicted the subsequent course of the tournament all the way to the finals.”
A large number of factors were included in the model: the bodily dimensions of the players; the number of players who reached the Champions League or European Cup finals in the previous season; whether or not there are groups of nominated players from the same club, who are presumably used to playing together; the strength of the respective national league; the number of players active in foreign countries and their level of play; position in the world ranking list; the valuation of the continental federation to which the respective nation belongs; national population and gross national product; the home-court advantages of the two host countries; age and tenure of the coach as well as the national origin of the coach compared to the country being coached.
Less coincidence than in soccer
Part of the team had already used this method to generate forecasts for the previous two FIFA Soccer World Cup competitions; however, as yet there have been no similar predictions for handball tournaments. “Coincidence plays a smaller role in handball than it does in soccer,” says Jörn Uhrmeister. “Since there are more countable events, i.e. more goals, as a rule the favorite comes out on top.” Accordingly, he says, is should be easier to predict a handball world champion than a soccer world champion. “However, ” Uhrmeister continues, “We of course still don't know which players will be injured, whether a given player will put in a particularly strong appearance or whether or not a given team might play its way into a frenzy.” Uhrmeister added that it is also possible that the method is oriented too strongly to individual variables.
According to Jörn Uhrmeister matters get particularly difficult when predicting the winners of contests between two European countries: “Handball is only popular in a few countries outside of Europe, so the continental federations outside of Europe play a subordinate role compared to the European federation,” he says. In the past only one non-European nation has managed to reach the winners' podium in a major tournament: Qatar took third place in the world championships held in its own country in 2015.
Once the final group of 16 had been announced, the researchers will repeat the analysis based on this data; the results will be available on the TU Dortmund Website starting on Monday, January 14, 2019
Dr. Gunther Schauberger
Technical University of Munich
Chair of Epidemiology