A prediction model that anyone can try has given Brazil a 25 per cent chance of winning the men’s football World Cup – but the results are sobering reading for many other nations
18 November 2022
Brazil are the most likely winners of the 2022 football World Cup according to a prediction model from the Alan Turing Institute in London. The publicly accessible model gives Brazil a 1-in-4 chance, with England’s chances put at less than 1 in 10.
Many people, from bookmakers to bankers, have run models trying to predict the winner of the 2022 men’s football World Cup in Qatar, but most of these are run behind closed doors.
Nick Barlow at the Alan Turing Institute and his colleagues have developed a model that people can run on their computers at home, with 1000 tournament run-throughs taking 15 minutes on an average laptop.
“It’s quite important to us for most of the things we do that we make them open source,” says Barlow. “We encourage people to get involved, to use our code and to contribute to it.”
When Barlow and his team ran the tournament through 100,000 times using their model, they found that Brazil won 25 per cent of the time, with their next-closest rivals being Belgium on 19 per cent and Argentina on 13 per cent.
The researchers adapted a common method used for matches in domestic leagues that gives teams a score for defence and attack to predict matches, but they tweaked their model to eliminate the home advantage that will be absent for all teams in Qatar apart from the home nation, as well as accounting for differences between the strength of teams that play each other in international friendlies.
They also tuned it to give more weight to the results of certain matches, like semifinals and finals, and more recent games, as well as running the model on past tournaments to see how well its predictions matched up to the real-world results, and tweaking it based on its performance.
Barlow’s model agrees with a model from Achim Zeileis at the University of Innsbruck in Austria and his colleagues, who ran their algorithm on a supercomputer to find that Brazil were also the likely winners, putting their chances at 15 per cent.
But other models predict different winners. Insurance company Lloyd’s used the collective insurable value of a team’s players to predict that England will win by beating Brazil in the final. The same model correctly predicted Germany to win the men’s World Cup in 2014 and France to win in 2018.
Whereas Belgium has been predicted as the most likely winner by a model designed by Matthew Penn at the University of Oxford and his colleagues that correctly predicted the men’s Euro 2020 winner as Italy and six of the eight quarter-finalists. This model assumes goals scored and conceded are evenly distributed around an average value.
More on these topics: