After watching the World Cup 100,000 times, AI predicts that Brazil will win the sixth championship trophy in team history, with a winning rate of 25%!

Big Data Digest Produced

Author: Caleb

The World Cup schedule has come to the 11th day.

I believe that Argentina fans were sweating last night. Although Argentina was under siege throughout the game and had a total of 23 shots, they missed too many opportunities, including Messi's penalty kick. But fortunately In the second half, the attack was rapid and they scored 1 goal in the opening game and then pursued the victory, eventually defeating the Polish team 2-0.

This game also enabled Messi to surpass Maradona and become the player with the most appearances in the World Cup in Argentina's history.

The group stage continued to be upset. In the final round of Group D, the Tunisian team defeated the French team, which started with a non-main lineup, 1:0, but ranked third in the group and missed the qualifying group. The French team advanced to the top 16 as the group leader.

Although the World Cup is not yet halfway through, the predictions about the winner of this World Cup have begun early.

Will it be the traditional strong team Brazil and France? Or is there a dark horse emerging?

Twelve years after Octopus Paul retired gloriously, Squidward may not have thought that his mantle would be inherited by AI.

Let’s take a look at how AI predicts the World Cup winner~

Optimize the model and database

The Alan Turing Institute (The Nick Barlow, Jack Roberts and Ryan Chan, three researchers at the Alan Turing Institute, developed a statistical model.

This statistical model is based on AIrsenal, a model derived from "Premier League Fantasy" in 2018, and AIrsenal draws on the classic football prediction model launched by Dixon and Coles in 1997.

The classic model mainly starts from the team's offensive strength, defensive strength and home court advantage, and uses Bayesian statistics to calculate the team most likely to score and win in the game.

In order to make AIrsenal more valuable in predicting the results of international competitions, researchers introduced the relative strength parameters of different continents.

#In addition, researchers also used a more comprehensive database when training the model.

The database compiled by GitHub user martj42 records detailed results of every football match since 1872. The data also differentiates and sorts the weight of the games. For example, the World Cup main matches (after 2002) and recent games are sequentially greater than the intercontinental championships, qualifiers and friendly matches.

GitHub link: https://github.com/martj42/international_results

The researchers also input FIFA official rankings into the model to provide a Latest estimates of team performance.

Who will win the championship this year?

After 100,000 simulations, the AI ​​also gave its own predictions.

Unsurprisingly, the Brazilian team has always been the favorite, reaching the top of this World Cup in 25,000 predictions, with a winning rate of 25%.

Followed by the traditional strong teams Belgium and Argentina, the winning rates reached 18% and 13.5%.

#If you think it’s still a bit far away to talk about the finals and the championship, let’s take a look at the predictions of recent game results.

As for the match between England and Wales on November 29, the result given by AI is that England has about 80% probability of qualifying from the group stage, and the probability of entering the quarter-finals or finals is is 60%, and the probability of winning the championship is only 7%.

Looking at Wales, the Welsh team has a 50% probability of qualifying for the group stage, but the probability of entering the final is only 2%, and the probability of winning the championship is only 0.5%.

Judging from the current performance of this World Cup, England has 1 win, 1 draw and 4 points, temporarily ranking first in the group. After easily defeating Iran 6-2 in the first round, their performance against the United States in the second round was mediocre. Generally speaking, after two rounds of games, England is still relatively smooth, and the physical exertion is not very high, but the problem is that the team's lineup has not yet faced a real challenge. In the last group match before the knockout rounds, England definitely needs more energy to face the game. Although the overall strength of Wales is dominated by Bale and Ramsey, the strength of the others is obviously not as good as before due to age and injuries.

At the same time, judging from the historical results of the two teams playing against each other 103 times, England has an absolute advantage with 68 wins, 21 draws and 14 losses.

Judging from the final result, England ended the group stage with a 3:0 victory. At the same time, they also ranked first in Group B with a record of 2 wins and 1 draw, joining hands with the United States team, which had 1 win and 2 draws. qualify.

The Brazilian team is very popular with AI

In addition to the model of the Alan Turing Institute, there are many prediction models that have given their own predictions for this World Cup Prediction.

For example, blogger @JackCuiAI counted the international football match data from 1872 to 2022 and the FIFA world ranking data from 1922 to 2022, analyzed these data, and extracted the factors that affect the game. After including key features such as ranking, home and away games, number of shots, number of saves, and final score, the Gradient Boosting Classifier uses the integrated learning gradient boosting decision tree to give its own prediction.

In the first four games of Qatar-Ecuador, England-Iran, Netherlands-Senegal and Wales-USA, the AI ​​gave predictions that were completely consistent with the actual results, even Including group stage scores and points for each team.

According to AI prediction, the top 8 of this World Cup will be the Netherlands, Argentina, Spain, Brazil, England, France, Belgium, and Portugal. The final will be between Brazil and Portugal, and Brazil will win the final. Won the sixth World Cup championship trophy in team history.

It seems that the Brazilian team is indeed very popular with AI.

The joint research team from TU Dortmund, Technical University of Munich, University of Luxembourg and other universities also combined multiple statistical models representing team strength with team structure (market value or number of Champions League players) and the society of the country of origin Information such as economic factors (population or GDP) are given for World Cup predictions.

Brazil tops the list with a 15% winning rate and a 23.5% probability of entering the finals; Argentina is closely followed with a 11.2% probability of winning; the third favorite is the Netherlands, with a winning rate of 9.7%; Germany , France ranked fourth and fifth.

The research team’s previous predictions have been very accurate, such as the European Cup champions in 2008 and 2012, the World Cup champions in 2010, etc.

But Grohl, a professor of statistics at the Technical University of Dortmund, also said that there may be errors in the predictions, otherwise the football match will become very boring.

"What we provide is only probability, not certainty. 15% probability of winning also means 85% probability of failure."

Yes, the uncertainty on the court is The most attractive one, right? Messi.

Related reports: https://www.turing.ac.uk/blog/can-our-algorithm-predict-winner-2022-football-world-cup

The above is the detailed content of After watching the World Cup 100,000 times, AI predicts that Brazil will win the sixth championship trophy in team history, with a winning rate of 25%!. For more information, please follow other related articles on the PHP Chinese website!