top of page
Search

Can AI Outperform Football Experts in Predicting Match Outcomes?

As a lifelong football fan and someone who loves technology, I’ve become intrigued by how Artificial Intelligence (AI) is making waves in the world of football, especially in predicting match outcomes. With the rise of advanced supercomputers and AI models designed for football predictions, I am intrigued by the question: who makes more accurate predictions — AI or experienced football fans? If you’re as curious as I am, keep reading to discover the answer.

First, let’s explore how AI makes predictions in football. AI makes football predictions by first ingesting a large dataset that includes historical match results, player statistics, injury reports, and other relevant information. The AI model is then trained on this data to identify patterns, trends, and factors influencing game outcomes. Once trained, the model can predict future match results by applying these learned patterns to new data. The accuracy of these predictions generally improves with the quantity and quality of the data provided, which is why some AI models perform better than others.

People often argue that AI predictions are superior because they can analyze vast amounts of data and detect complex patterns that humans might miss. However, others contend that football fans might still offer better predictions since AI models may overlook crucial external factors, such as tactical changes and team morale, that can impact a match right before it begins.

Since I was curious about whether AI or humans make better predictions for football games, I looked for studies comparing the two, but I couldn’t find any. Consequently, I decided to design my own experiment to determine which approach is more accurate.

Here’s an overview of my experiment: I will compare predictions for Euro 2024 from the OPTA supercomputer against those from top football pundits. Points will be awarded based on prediction accuracy as follows: 15 points for correctly predicting a quarter-finalist, 20 points for a semi-finalist, 25 points for a finalist, and 30 points for predicting the winner. I will average the points earned by the pundits and compare these results with the OPTA supercomputer’s performance to reveal once and for all who is better at predicting football outcomes.

This is how the OPTA supercomputer works: it’s a leading AI tool for football predictions. By simulating Euro 2024 a remarkable 10,000 times, it uses advanced algorithms to analyze extensive data and various scenarios to provide highly accurate predictions. For comparison, the predictions from football pundits were sourced from a well-regarded ESPN YouTube video, known for featuring some of the top football experts.

The actual euro 2024 predictions from the quarter finals stage onwards is shown below:

Let’s start by seeing what the OPTA supercomputer predicted. The predictions from the supercomputer are displayed below.

The key insights from the supercomputer’s predictions are as follows: it accurately predicted 1 of the finalists and 3 out of 4 semi-finalists. Although it missed the winner, it correctly identified 6 out of 8 quarter-finalists. Overall, these results reflect impressive and solid predictions. The supercomputer earned 175 out of a possible 280 points in the experiment. That is significantly higher than I expected.

On the other hand, let’s examine how the pundits performed. Many of the pundits shared similar predictions, frequently including England, France, and Germany as semifinalists (with England and France actually making it to the semifinals). Unlike the OPTA supercomputer, several pundits accurately predicted Switzerland reaching the quarter-finals. Now, let’s review the points each pundit earned. Below are the names of the pundits along with their respective scores and the average score.

Marc Ogden: 125 points

Stewart Robson: 190 points

Steve Nicol: 190 points

Average score: 168.3 points

It looks like the OPTA supercomputer emerged as the overall winner! Although 2 out of the 3 pundits outperformed the supercomputer individually, the average score of the pundits was about 8 points lower. Despite this, it could be argued that the pundits technically won since the majority (2 of 3) surpassed the computer’s performance. From this, we can see that the OPTA supercomputer generally makes better predictions than the average football fan, but it doesn’t always outperform every expert.


References

Myson, C. (2024, June 21). Who will win Euro 2024? The Opta predictions. The Analyst. https://theanalyst.com/eu/2024/06/who-will-win-euro-2024-predictions-opta/

Matches (By UEFA). (2024). UEFA Euro 2024. Retrieved July 19, 2024, from https://www.uefa.com/euro2024/fixtures-results/bracket/

ESPN FC. (2024, June 27). PREDICTING THE EUROS 🔮 The crew makes EURO 2024 bracket picks | ESPN FC [Video]. YouTube. https://www.youtube.com/watch?v=QyuAXhe368Y



17 views0 comments

Comments


bottom of page