How does the supercomputer that correctly predicted the last three matches of Euro 2024 work?

Opta is a supercomputer familiar to sports fans across many tournaments. Although it is not always accurate due to many unpredictable factors in each match, Opta is described as "the analyst who creates the big surprises". This supercomputer has also become a reference source for analysis and prediction of results at Euro 2024.

Before Euro 2024, Opta predicted that England would be crowned champions. However, throughout the tournament, the supercomputer was continuously updated with parameters of the teams playing to conduct analysis, from which it predicted that Spain and England would advance to the final and the cup would belong to Spain.

How Opta predicts

Opta Sports, based in London, has been analysing football matches since 1996. CEO Aiden Cooney is not considered a sports expert but has a keen eye for statistics and has turned "soulless" numbers into valuable commodities.

Sharing on Twitter in 2018, Cooney said he had the idea of ​​collecting player stats "NASA style", such as data from player tracking chips, satellites measuring the distance players run in 90 minutes, looking at the direction of attack, heat maps when players move. They also use laptops, super-powerful PCs, huge data storage servers and protect it "like a fortress".

Opta launched its real-time data collection process for football matches in 2006. The company then continued to expand to tournaments in other sports.

Simon Banoub, Opta's chief marketing officer, told Stuff the company compiles more than 2,000 pieces of data per match, covering more than 60,000 matches a year across 30 sports in 70 countries, providing results to websites, newspapers, broadcasters and teams.

“We have multiple data collection centers around the world. During each match, we have three people using custom software watching live on the field: one person in charge of the home team, one person in charge of the away team, and a third person who feeds this data back to the supercomputer,” said Banoub.

For less important matches, analysis is done via video or data from on-field staff, according to Banoub. But everyone is trained to understand a tackle or a duel, and to know the difference between a short pass and a goal kick.

“People hired at Opta need two main skills: being good at video games and knowing about football,” says Banoub, who says that’s why the company hires a large number of students as analysts.

Before being fed into the supercomputer, the data is processed raw. The numbers are coded by experts according to the player's actions, meaning shots, passes, saves, headers, crosses... are all classified. "It's not as simple as just sitting and watching football. Therefore, out of 100 people interviewed, only 5 pass the initial training process," Banoub said.

Opta also uses computer vision technology to automatically track the movements and actions of specific players during a match. By analyzing the video feed, computer algorithms can identify and extract data points such as player position, running distance and passing accuracy.

In addition, Opta uses technology such as GPS devices, accelerometers and other wearables to collect player performance data during training and matches, providing insights into fitness, speed, distance covered and other performance metrics.

The company currently partners with many sports leagues and organizations to receive official data such as statistics, player information, and other relevant details.

Finally, the data will be aggregated and fed into a supercomputer to run a simulation of the hypothetical match about 10,000 times, then take the average value. However, the data is constantly changing before and during the tournament, so the results will also be adjusted regularly.

According to Metro, unlike other supercomputers that typically process all data on the system, Opta Sports uses a combination of human expertise, advanced technology and partnerships to collect and analyze, ultimately producing valuable reference statistics for sports teams, media and fans around the world.

Cannot be absolutely accurate

Despite relying on a wealth of statistical data, Opta doesn't always get its predictions right. This is obvious because there are always unexpected factors that arise throughout the season that the supercomputer can't predict.

For example, when predicting the 2022-2023 Premier League champions, Opta gave Liverpool a 49.7% chance of winning the trophy, while Manchester City had a 47% chance. As a result, Liverpool finished 5th, while Manchester City won the title.

At Euro 2024, before the tournament, Opta predicted England had a 48.2% chance of reaching the semi-finals, a 31.1% chance of reaching the final and a 19.9% ​​chance of winning the trophy. Spain were fourth with a 9.6% chance, shown in purple in the chart above.

However, after updating the round of 16 data, the supercomputer revised its prediction, putting Spain in first place, with England in second (red). In fact, both teams advanced to the final, with Spain winning.

Opta's accuracy rate is "up to 98%", Mr Banoub said. "There is always a combination of objective and subjective factors, such as motivation, how players help their teammates, how they react when they are in trouble," he said.

However, Dr Alireza Monajati, a UK-based data scientist, said the 98% rate was unlikely to be credible. "They cannot guarantee such accuracy," Monajati told the BBC. He also assessed Opta as a reference channel but needed to compare it with other sources, instead of completely trusting it.

In addition to football, Opta also provides data for other sports around the world. The company's biggest rival is ProZone, which provides even more detailed prediction data than Opta, but does not make it public and only sells it to large corporations.