We never told our model who the favourites were. We just handed it player data, the kind any scout can pull, market values, recent form, international minutes, and asked it to build each of the 48 World Cup teams from the players upward. No reputations, no seedings, no peeking at what anyone else predicted. Then we ran the whole tournament through a Monte Carlo simulation five million times over and let the bracket fight it out. (A Monte Carlo simulation is just the grown-up version of playing out the tournament on paper, except you do it five million times so a single lucky bounce never gets the last word.)
It put France on top. Just barely. And, because we are Belgian and could not help ourselves, the very first number we went looking for was a red one.
So, about Belgium
The Red Devils land ninth on our board, with a 2.8% shot at the trophy and a 16% chance of reaching the last four. We will be honest, we hoped for higher. But ninth out of 48 is exactly where a squad like this should sit: good enough to ruin anyone’s afternoon over ninety minutes, not deep enough to be favoured across seven straight knockouts. The model is not being mean. It is being a scout, and scouts do not get sentimental about the shirt.
The upside, if you squint: a 16% run to the semis is a coin-flip away from a coin-flip, and weirder things have happened in a Belgian summer. And here is the stat we are taping to the office wall: across our five million simulated tournaments, Belgium lifted the trophy in roughly 140,000 of them. A hundred and forty thousand parallel universes where it all comes good. I am choosing to live in one of those universes.
A favourite, but only just
At the top, France comes out as the single most likely champion at 15.9%, which sounds decisive right up until you notice who is breathing down its neck. Spain and England sit at 13.8% and 13.6%, close enough that the gap is basically a rounding error. Portugal trails at 10.3%, and then a pack of heavyweights, Brazil, Germany and Argentina, bunch up between 7% and 8%.
So the real headline is not France. It is that nobody in this tournament is better than a one-in-six bet to win it. The top four are separated by a couple of percentage points, with four more genuine contenders queued up behind them. On the numbers, 2026 is the most wide-open World Cup in years, which is either thrilling or stressful depending on who you support.
Upset watch. If you like a bandwagon, our model has a soft spot for Morocco, who reach the quarter-finals more than one time in five and have the look of a team nobody wants to draw. Host nation USA and an unfashionably solid Switzerland round out the list of sides priced to ruin a favourite’s tournament. You have been warned.
| Team | Reach semis | Reach final | Win it |
|---|---|---|---|
| France | 38.6% | 25.3% | 15.9% |
| Spain | 37.4% | 22.9% | 13.8% |
| England | 36.7% | 23.2% | 13.6% |
| Portugal | 33.2% | 18.8% | 10.3% |
| Brazil | 25.4% | 14.8% | 8.0% |
| Germany | 26.2% | 14.9% | 7.9% |
| Argentina | 26.2% | 13.8% | 7.0% |
| Netherlands | 20.6% | 11.0% | 5.3% |
| Belgium | 16.1% | 6.9% | 2.8% |
For the record, our model rates Belgium ahead of our neighbours’ optimism and just behind the Netherlands, who edge it at 5.3%. We are choosing not to dwell on that one.
Where we disagree with the crowd
The fun part of any model is not where it nods along with everyone else but where it stubbornly does not. Two teams stand out. We rate England a clear notch higher than the prediction markets do, 13.6% against the crowd’s 9.8%, and Germany too, 7.9% against 5.3%. In both cases the player data is simply richer than recent results suggest, and our model trusts the team sheet over the mood.
These are not glitches. They are our model quietly pointing out that the talent on the pitch has not yet shown up on the scoreboard, which is exactly the kind of gap a recruiter dreams about finding before everyone else does.
The bit that should not have worked
Here is the test that matters, and the one that genuinely surprised us. At no point did we show our model what anyone else expected. The prediction markets, where thousands of people back their own World Cup calls with real money, were kept completely out of sight. We built the teams from player data, ran the simulation, and only afterwards laid our numbers next to the crowd’s.
They agreed almost exactly. The correlation between our title probabilities and the prediction markets was 0.975, which is about as close as two independent guesses at the same thing ever get. Our model had only ever looked at players, never at a single prediction, and it reconstructed the collective wisdom of thousands of people on its own.
That is the whole point, and we will spare you the inner workings. What it proves is the thing we keep coming back to: read player data carefully enough and it carries the real signal. The strength of a team, the gap between two sides, the slow burn of a squad whose results have not caught up to its talent, it is all sitting in the player data if you know how to read it. Thousands of people, working hard at their predictions, landed in the same place a pile of player stats did.
The same engine that reads a squad for fit
Predicting a World Cup is a genuinely fun way to pressure-test that idea against a hard, public benchmark, and a good excuse to argue about Belgium’s chances at the office. But it is the same machinery we point at the problem clubs actually pay us to solve. Turning raw player data into an honest read on quality, fit and gaps is what gaffer does, whether the question is who lifts the trophy in New York next summer or who fits the side you already have, before you spend the fee finding out the hard way.
And no, before you ask, we did not rig it for Belgium. We checked. Twice. The model just likes France.