Why use experts like Paul the octopus when we can use science and informatics? On the side of the Alan Turing Institute in London, explains New Scientista few football and algorithm lovers have designed a program capable of simulating the matches of the 2022 World Cup a large number of times. A quarter of an hour is enough to run a thousand different tournaments on a simple computer. living room.
The simulation in question being available for free onlineit is therefore possible for everyone to organize their own tournament, or even to examine the way in which the program has been designed. “It is important for us to publish our software in open source”explains Nick Barlow, of the Alan Turing Institute. “We encourage people to get involved, use our code and contribute to it.”
Nick Barlow’s team was based on AIrsenala method she had developed in 2018 to be able to play “fantasy soccer” in the English Premier League. This was based itself on the work by Dixon and Coleswhich serve as a reference for football predictions, but it also includes Bayesian statisticswhich includes a degree of belief in the probabilities.
On his site, the study group explains that it has tried to deal with the biases most likely to negatively influence its work. For example, Brazil having not met a European team since 2019, it is very difficult to assess what the face-to-face between the Seleção and opponents like France, Belgium or England could give. .
The algorithm, to which the scientists provided the results of the direct oppositions between the thirty-two teams present in Qatar, was also trained using data from the last five world tournaments, the oldest World Cup taken into account. being the last won to date by Brazil, in 2002.
The FIFA Ranking was also taken into account, which is not the case of the individual characteristics of the players, nor their success rate in penalties, nor the place of the tournament or the weather conditions.
One in four chance for the Seleção
After simulating no less than 100,000 virtual World Cups, British scientists delivered their results: in 25% of cases, Brazil won. Belgium is also well placed with 19% wins, ahead of Argentina with 13%. France is fourth, with an 11% chance.
According to New Scientist, which relays the results of this giant simulation, studies of the same ilk also give Brazil the winner of its sixth World Cup, but this announcement is far from unanimous. On the side of the insurance company Lloyd’s, for example, it is said that England will beat Brazil in the final, supporting figures. However, the model used in this last study had made it possible to predict the victory of Germany in 2014 and that of France in 2018.
As for the modeling carried out at Oxford University by Matthew Penn’s team, which had rightly given Italy the winner of Euro 2020 and announced six of the eight quarter-finalists of the tournament, it indicates that Belgium is the most likely winner.
According to the very complete figures provided by the Alan Turing Institute, you will also learn that the France team has an 87.4% chance of going to the round of 16, 60% of playing the quarters, 40.6% to be a semi-finalist, and 21.7% to reach the final. Each time, the Blues rank fourth in terms of probabilities, behind Argentina, Brazil and Belgium – in order. There you go, you know everything, and so you can boycott the world cup peace of mind.