The Brazilian national team will play its first match at the World Cup in Qatar this Thursday (24). And, if it depends on the predictions of an artificial intelligence developed by the retailer Magazine Luiza, Brazil will reach the final together with France, winner of the last world championship, in 2018.
The company’s data scientists analyzed 10,000 different scenarios to arrive at the result. According to the algorithmic model, Brazil has a 19% chance of winning the 2022 Cup, while the French team has a 16%.
The statistics of “Bolão da Lu do Magalu” are close to the predictions made by an international team of researchers from the Universities of Innsbruck (Austria), Ghent (Belgium), Luxembourg (Luxembourg) and the Technical Universities of Dortmund and Munich (both in Germany ).
According to the model developed by the group above, Brazil is the biggest favorite this year, with a 15% chance of taking the FIFA trophy, followed by France, Germany, Holland and Argentina.
How artificial intelligence arrived at the result
To arrive at the proposed results, the data science team from the retailer’s technology area, Luizalabs, used artificial intelligence tools, machine learning (system learning capacity) and neural networks, fed with information from the last 30 years (from 1992 until now) of world football.
According to the professionals, the model used data from the history of games from that period, cup scores, players’ lineup notes – goalkeeper, defense, midfielder and attack – and from the updated Fifa ranking, which classifies countries according to the teams’ performance.
Each of the 10,000 defined scenarios featured a champion. Brazil won 1,900 times, followed by France, which appeared as champion 1,600 times. The team of international researchers performed 100,000 simulations, and reached similar results.
“Football is a very difficult sport to predict results and that’s also what makes it so exciting. Sometimes it’s a detail that defines the result of the game. There are teams that have a great attack, but don’t have a good defense , for example. The variables take into account game by game and each one has the probability of winning, drawing or losing”, explains Fernando Nagano, director of analytics at Magalu, with exclusivity to Tilt🇧🇷
AI guessed winner of match between France and Australia
Following Magalu’s tool, in the game France and Australia last Tuesday (22), Mbappé’s team had a 60% chance of winning, 22% of a draw and 18% of defeat. The official result confirmed the prediction, with 4 goals for France and zero for Australia.
At the end of the first round of the championship, the model will run new simulations, based on the results of those classified in each phase, updating the probabilities.
AI wrong prediction between Saudi Arabia and Argentina
“Since it’s about probabilities, sometimes it can give an upset. At the start of the championship, Saudi Arabia had less chances than Argentina, according to our model. But with the result of the last score [Arábia Saudita 2 x Argentina 1]the chance of the Latin American country will reduce”, says Mangano.
He also recalled the Croatia national team in the 2018 World Cup, which in the AI and statistics models of the time was not one of the favorites, but managed to reach the final, losing to France.
According to the director, the model was developed in just two weeks.
“In our routine, we build predictive models to support the company’s decision-making. We took advantage of this knowledge to develop this game for the World Cup. Even so, we had to do a lot of research because we are specialists in retail and not in football”, he joked.
Can help with bubbles
The tool developed by Magalu will publish, as of this Thursday (24), the guess on the result of each game on the company’s social networks.
The proposal is that Internet users can use the predictions to make their sweepstakes from computer-processed data.