A group of European statisticians claim to have built a machine learning algorithm that can predict the winner of this year’s FIFA World Cup.
The model works by processing a variety of data about national teams and players before running 100,000 simulations to find the most likely winner.
Using this method, the team’s lead author, Dr. Achim Zeirais, a statistician at the University of Innsbruck, explained in a blog post that Spain has the best chance of winning this year’s competition, followed by England.
This is not the first time the group, which includes researchers from Germany’s Dortmund University of Technology and Munich University of Technology, Norway’s Molde University, and Austria’s University of Innsbruck, has attempted to use machine learning to predict the World Cup winner.
In 2018, the team correctly predicted the United States as the winner of the 2019 Women’s World Cup, but was less fortunate, choosing Spain and Argentina as the winners of the 2023 Women’s and 2022 Men’s tournaments, respectively.
The algorithm that predicts this year’s World Cup works by processing available data on all domestic matches from the past eight years.
This is combined with ‘expected’ strength estimates for each team provided by combined odds data obtained from international bookmakers.
The overall strength of each team will be fine-tuned depending on each player’s rating. The rating is derived from data on performance at both club and international level and expectations in the international transfer market.
This data is ultimately fed into a “Random Forest” machine learning algorithm that evaluates each team’s chances of winning in various matchups.
While confident in the model’s educated guesses about this year’s winner, Zairais warned that it was just a prediction and was hostage to statistical shocks like the recent goalless draw between Cape Verde and Spain.
“All of our predictions are probabilistic, obviously less than 100%, and therefore by no means certain,” Zeileis wrote. “While we can quantify this uncertainty in terms of probabilities from different tournament possibilities, which of these potential tournaments we will ultimately see during the actual tournament is not determined in advance.”
This is not the first serious attempt to predict the World Cup winner. In 2014, German mathematician Joachim Clement used a statistical model to accurately predict the national team’s victory in that year’s tournament, as well as the winners of three subsequent competitions.
This followed his extremely accurate match-by-match predictions for Euro 2008, followed by Octopus (also German) correctly predicting Spain as the winners of the 2010 tournament, a streak that unfortunately ended with the death of the cephalopod three months into the tournament.
read more:
