Spain tops World Cup forecast as U.S. title odds sit at 1%
A statistics team ran 100,000 tournament simulations and made Spain the narrow favorite, with the U.S. likely to advance but unlikely to win.
By Hana Yoshida · Markets Reporter
3 min read
A machine-learning forecast led by statistician Achim Zeileis gives the United States a 1% chance of winning the 2026 World Cup on home soil. The projection matters because it shows a favorable early path for the U.S. men’s team but a steep drop in its odds once knockout matches begin.
Zeileis, a professor of statistics at the University of Innsbruck, wrote in The Conversation that his group simulated the tournament 100,000 times after estimating team strengths from match data, betting markets, player ratings and other factors. The model makes Spain the leading title candidate at 14.5%, ahead of England and France at 12.4% each and Germany at 11.2%.
The forecast puts Portugal at 8.9% and Argentina at 8.2%, according to Zeileis. He said the expanded 48-team format and five knockout rounds leave the leading contenders bunched more tightly than in smaller tournaments.
U.S. favored to advance from its group
The United States has a 78% chance of reaching the Round of 32, the highest probability among the four teams in its group, according to Zeileis. After that, the model shows the U.S. survival odds falling quickly because each match becomes an elimination game.
The final is scheduled for July 19 at MetLife Stadium in New Jersey, according to the forecast write-up. Zeileis said the model gives the U.S. a 1% chance of winning the title there.
The system also produces match-level probabilities. For the opening match, Zeileis said Mexico averages 1.9 goals in the model, compared with 0.7 for South Africa. That translates into a 65% chance of a Mexico win, a 21% chance of a draw and a 14% chance of a South Africa win, according to the forecast.
How the model was built
Zeileis said the research team first estimated team and player strength, then used machine learning to combine those estimates with other team and country data. The tournament simulations followed the official draw and FIFA rules, including extra time and penalty shootouts.
The inputs included national-team matches from the previous eight years and betting odds from international bookmakers, according to Zeileis. The team also used player ratings based on goal contributions for clubs and national teams, plus expected market values drawn from Transfermarkt.
Zeileis said the model considered additional indicators, including FIFA ranking, how many players a team had in the Champions League semifinals, and socioeconomic measures such as GDP per capita. A random forest model, trained on major tournament matches since the 2006 World Cup, connected those factors to expected goal totals.
The forecasting group includes Andreas Groll and Rouven Michels and colleagues at TU Dortmund University, Lars Magnus Hvattum of Molde University College, Gunther Schauberger of TU Munich and Zeileis, according to The Conversation.
Zeileis said the group correctly picked the United States to win the 2019 Women’s World Cup. For the 2023 Women’s World Cup and the 2022 men’s World Cup, he said the eventual winners, Spain and Argentina, were treated as serious contenders but were not the model’s favorites.
Zeileis cautioned that the forecast is probabilistic, not a certainty. The model ranks likely outcomes, but the tournament result still depends on matches that can swing on draws, extra time and penalties.
This story draws on original reporting from Fortune.