The 2026 World Cup kicks off on June 11, and the predictions are already louder than the qualifiers ever were. Forty-eight teams, three host countries, the biggest tournament the sport has ever staged. Everyone has a pick. Your group chat has a pick. And the most-quoted pick of all this year comes from a machine.
Opta’s supercomputer ran the whole tournament 10,000 times and came back with a favorite: Spain, at 16.1 percent. That number has been everywhere for the past week. What almost nobody is talking about is how thin that lead actually is, which teams the model quietly rates far lower than you’d expect, and the one big thing I think the simulation is getting wrong.
Let’s get into it.
Here’s a wrinkle that’s genuinely useful if you follow the odds.
On a couple of teams, the model and the betting market flatly disagree. Both Brazil and Portugal are rated noticeably lower by Opta than they’re priced by the bookmakers. That’s a real divergence, and it’s worth understanding what it means rather than just picking a side.
A model like this works from underlying numbers: how teams have actually performed, match after match, stripped of reputation. A betting market works from those numbers too, but it’s also pricing in public money, brand-name pull, and the simple fact that a lot of people will always back Brazil. When the two disagree, it usually means the market is leaning on history and hype that the data doesn’t fully support. That doesn’t make the model right and the bookies wrong, but it tells you where the “obvious” pick might be softer than it looks. (None of this is betting advice. It’s just an interesting place to point your skepticism.)
