When Achim Zeileis and his team of statisticians ran their machine learning model through 100,000 simulations of the 2026 World Cup, they uncovered something both thrilling and sobering: no single nation dominates the odds. Spain edges ahead as the favorite at just 14.5%, a margin so thin that the tournament is shaping up to be one of the most competitive in recent memory.
The forecast, developed by researchers at the University of Innsbruck and TU Dortmund University with collaborators from the Technical University of Munich and Molde University College in Norway, reflects the genuine uncertainty facing the sport's elite teams as they prepare for the tournament in Canada, Mexico, and the United States. England and France both sit at 12.4%, while Germany trails only slightly at 11.2%. Even further down the ladder, traditional powerhouses like Portugal (8.9%), Argentina (8.2%), the Netherlands (5.6%), and Brazil (4.7%) remain in contention. As Zeileis observes, "Compared to previous tournaments, this year's title race is very tight."
Building this forecast required wrestling with mountains of data and methodological complexity. The researchers synthesized team performance from past international matches, current bookmaker odds, player ratings culled from both club and international competition, and the average market value of each squad. Rather than relying on any single indicator, they fed all this information into a machine learning algorithm designed to identify hidden patterns across the entire dataset. But the work came with unexpected obstacles. "On the one hand, we had to compile all this data, some of which is only available very shortly before the tournament," Zeileis explains. "For example, we've only known the final rosters of all 48 teams for a few days."
The team then executed an ambitious validation strategy: simulating the entire tournament game by game, 100,000 times over, while respecting the actual tournament draw and all FIFA regulations. Rouven Michels, working from TU Dortmund's team, describes the delicate balance required: "The challenge was also to combine statistical expertise and machine learning in such a way that a robust model of the tournament could be built."
The researchers' track record offers some reassurance. In previous predictions, their favored teams have gone on to win several times—the 2010 World Cup, Euro 2012, and the 2019 Women's World Cup among them. Yet Andreas Groll cautions against reading too much into any single outcome. "The probability that the top favorite will actually win the tournament is usually no more than 20%, which conversely also means that some other team wins with a probability of 80%," he notes. "As a statistician, I'm therefore more interested in whether, on average, many of the teams we predict to go far will actually do so."
For Zeileis, the deeper value lies not in predicting champions, but in how the forecast itself can illuminate probability for millions of fans. "A tournament like this is a wonderful opportunity to spark an interest in probability among a huge number of people who would otherwise not come into contact with it," he reflects. It's a reminder that World Cup forecasting, at its best, is less about certainty than about capturing the beautiful complexity of sport itself—where Spain's 14.5% advantage leaves plenty of room for magic.