Even before the kickoff on June 11, the tournament had already been played tens of thousands of times. Not on a pitch, but on servers. Algorithms simulated every match, every group stage, every possible final — and delivered their verdict: Spain. But here is the paradox that no one really mentions: the model that crowns the Roja also gives it roughly an 84% chance of not winning.
Welcome to the most mathematically complex World Cup in history, with 48 teams for the first time. More teams means more matches, more possible scenarios, and — mechanically — favorites with lower probabilities. This is precisely what Chris Myson points out in the analysis published by Opta Analyst: "In a 48-team bracket, there is a huge range of possible outcomes, and no team is going to reach very high percentages."
25,000 finals in advance
Opta's supercomputer — the analytics division of Stats Perform, the world's leading name in sports data — simulated the entire tournament 25,000 times before the kickoff on June 11. Not 10,000 times as some media outlets reported: 25,000 complete simulations, match by match, from the first round to the July 19 final. The result of these simulations, published on June 1, 2026 according to Opta Analyst, is clear: Spain wins the tournament in 16.1% of scenarios, ahead of France (13.0%), England (11.2%) and Argentina (10.4%).
What these figures mean in practice: if you played the 2026 World Cup a hundred times in a row in a perfect simulator, Spain would win about sixteen times. France, thirteen. And in about 36% of cases, according to Opta, a country would become world champion for the very first time in its history.
Spain also shows another statistical quirk: it is the only team in the tournament to exceed the 50% threshold for reaching the quarterfinals, with exactly 52.1% according to Opta. For the final, the probability drops to 25.6%. In other words: even the top favorite has a three-in-four chance of not reaching the final.
The quiet genius of Siméon-Denis Poisson
To understand how an algorithm "simulates" a football match, we need to go back to a mathematical idea two centuries old. In the 19th century, the French mathematician Siméon-Denis Poisson was interested in the probability of rare events: how many times does an unlikely event occur within a given interval of time? His law — the Poisson distribution — answers this question with remarkable elegance.
A goal in football is precisely this type of event: rare, discrete (you don't score 2.7 goals), and relatively independent from one minute to the next. If a team scores an average of 1.8 goals per match, the Poisson distribution can be used to calculate the exact probability that it scores 0, 1, 2, 3 or more in a given match. This mathematical framework was first applied to football by statistician M. J. Maher in a landmark paper published in 1982, which showed that "an independent Poisson model gives a reasonably accurate description of football scores".
The central idea: each team has an offensive intensity parameter (let's call it λ, "lambda") that depends on its attacking strength and the opposing defense. A score is drawn for team A according to its Poisson distribution, a score for team B according to its own — and you have a simulated result. Repeat the operation for every match in the tournament, then do it again 25,000 times, and you get a probability distribution over all possible winners.
But a football match is not the sum of two independent dice rolls. When a team leads 1-0, both teams change tactics, which alters the goal intensities. This is why modern models — like the one described by Dixon and Coles in the Journal of the Royal Statistical Society — correct the naive Poisson model by weighting recent matches more heavily and adjusting the probabilities of low scores (0-0, 1-0, 0-1, 1-1), which behave differently from what an overly simple model predicts.
Monte Carlo at the stadium
Simulating a match is one thing. Simulating an entire tournament is another matter. This is where the Monte Carlo method comes in — a name borrowed from the Monegasque casino, which says a lot about the mindset behind it. The principle: repeat a random process a very large number of times to estimate the probability of an overall outcome. For a 48-team World Cup, that means simulating 104 matches per tournament, propagating the uncertainties from each round into the next.
This is exactly what Andreas Groll, Gunther Schauberger and Gerhard Tutz describe in their reference paper on predicting football tournaments: "The FIFA World Cup is repeatedly simulated and win probabilities are obtained for all teams." The model incorporates a regularized Poisson regression — a technique that avoids overfitting the data by penalizing too many variables — to estimate goal intensities from dozens of covariates: FIFA ranking, squad market value, recent form, average player age.
The University of Innsbruck team, led by Achim Zeileis, ran an independent simulation of 100,000 complete tournaments, incorporating final squads, bookmakers' odds and market values. The result converges, but differs slightly: Spain at 14.5%, England and France at 12.4% each. As Andreas Groll, a statistician at TU Dortmund, points out: "the probability that the top favorite actually wins the tournament generally does not exceed 20%", according to the University of Innsbruck press release.
Do you see the paradox? Three independent models — Opta, Innsbruck, and the University of Liverpool model, which gives Spain 26.1% according to its research release — agree on the name of the favorite but diverge on how strong a favorite it is. That alone is a lesson in statistics: agreement on the ranking does not guarantee agreement on the numbers.
The Haaland wildcard and the limits of models
There is a known flaw in all these models: they reason in terms of team averages, whereas knockout football is often decided by exceptional individual performances. Benjamin Holmes, a researcher at the University of Liverpool's Centre for Sports Business, explains this clearly: his model now simulates "injuries, suspensions and who scores the goals", even factoring in weather conditions and the altitude of North American stadiums.
This is where Erling Haaland enters the equation. A striker with a statistically outlying conversion rate — what data scientists sometimes call a "super-athlete" — can tip a team's odds in knockout matches, where a single goal can decide everything. Norway, not especially fearsome in the group stage according to the models, becomes a disruptive variable in single-match knockout simulations.
And then there is what no model can capture: a red card in the first minute, a gust of wind that deflects a shot onto the post, a goalkeeper who produces the match of his life. The Poisson distribution models the average intensity of goals — it does not model chaos.
That may be the real mathematical lesson of this exercise: a probabilistic model does not predict the future. It maps uncertainty. And on that map, Spain sits at the top — but the territory that eludes it remains, by far, the largest.
