Understand how websites define the chances of a team being champion or relegated

Will Atlético-MG end its 50-year fast without being Brazilian champion? Will the Libertadores trio will be with Palmeiras or Flamengo? Will Santos be relegated? Uncertainty occupies fans’ minds and should still last. But it is possible to approach the answers from a branch of mathematics: probability, a subject explored by websites, accompanied by fanatics, but not always understood.

Probability provides tools for complicated and useful everyday problems. “These are questions whose answers we cannot know precisely. In part, because they depend on a huge set of unknown variables. However, we can create mathematical models that help estimate that the probability of rain in São Paulo is 80%, making the residents go out with umbrellas,” says Marcelo Hilário, deputy coordinator of the Mathematics course at UFMG (Federal University of Minas Gerais) and a researcher of probabilities.

Click here and simulate the matches of the Brazilian Championship

The Chance de Goal, Infobola and Probabilities no Football websites provide estimates, with differences between them and variations at each round. They face a challenge: dealing with an extraordinary amount of information to predict outcomes, but needing to choose some variables to assess probabilities, simulating simplified models of a championship.

Here, nuances yield different results. “Simplifying too much can lead to a model that is not realistic enough; simplifying too little will lead to computationally intractable models,” says Hilário. “The task of mathematicians is to generate simple models to be simulated with our computational capacity, but which retain enough information to provide useful estimates,” he adds.

Chance de Gol defines each club’s strength parameters based on the results of the last 12 months, with decreasing weight according to the game’s longevity. The equation that simulates the results also uses the field command, with a fixed weight.

Marcelo Arruda, PhD in statistics and owner of Chance de Gol, explains that all national club competition games, including state ones, have weight in the model for the Brazilian Nationals. The site is the only one to measure odds for the knockout tournaments and to use each team’s strength parameter to estimate the odds before the start of competitions.

In international disputes, Chance de Gol only takes into account the results of this type of tournament – ​​and in the last 24 months. And without an extensive database, it doesn’t publish odds before the championship starts. “In the cups, I hope teams have a full round of the group stage to publish,” he justifies.

It is a different model compared to Tristão Garcia at Infobola and the UFMG Mathematics Department, at Probabilis no Futebol, which are similar, but with subtleties in the algorithms. In common, Tristão and the UFMG professionals define profiles for the clubs as hosts and visitors, based on their results in the tournament.

“The moment isn’t the last game, it’s an average. And that’s what prevails,” says Tristao.

“If you start winning, the odds of winning the next game increase,” defends Bernardo de Lima, a member of the UFMG project team and a researcher in probabilities.

In Infobola, the weight is total for the last six games and it goes down – 70% for the six previous ones, 50% for the following ones and 35% for the others. UFMG, on the other hand, uses a 5/6 geometric progression in which the weight of previous games decreases over the rounds.

With the probabilities of the models published on the websites, which they do not do before the 5th round of the Nacional, there is the challenge of dealing with the interpretation of the fans. “No matter how good the model, football is richer. Sometimes the numbers get in the way, the teams get scared. But the owner of the destination is the club”, assesses Tristão.

The professor at UFMG warns that a low probability will happen one day, as happened with Fluminense in 2009, when he avoided a relegation that had a 99% chance of materializing. “The probability is very high that something with a 5% chance will happen at some point,” says Bernardo.

The idea of ​​Infobola came to Tristão, in the 1999 Brazilian Nationals, when the average of the points of the competitions of that year and 1998 determined the relegation. With this experience, it gives tips for reading the numbers.

“50% probability of falling is a lot, but for a title it is little, as it is the pursuit of excellence. For the cup, from 1/3 of the championship onwards it only decreases the candidates. Relegation, on the other hand, is a 2nd round tournament” , he claims.

In addition to teaching the public to read the numbers, the way can be to help them get closer to mathematics and create a model, as professors at UFMG encouraged in open courses.

“One objective is to be a vehicle for scientific dissemination. We have already given short courses at universities, teaching probability, with the Law of Large Numbers in football. The idea was: ‘we are teaching you, if you don’t agree, do your algorithm'”, he says Bernardo

“The impartiality of the program [algoritimo] it is the greatest merit, since we are all partial”, concludes Tristao.