A 48-Team World Cup Has No Real Favourite. We Simulated It 10,000 Times to Prove It.
The expanded format has quietly killed the runaway favourite. Here is what the maths actually says.
Almost every World Cup preview you will read between now and kick-off is a feeling with a word count. A writer likes the look of Spain. Another one trusts the France squad. The Argentina pieces more or less type themselves around Messi. None of it is wrong, exactly, but none of it tells you anything you could not have guessed yourself.
We took a different route. Instead of picking a winner, we built a machine to play the whole tournament out for us, over and over, and counted how often each team came out on top.
The engine lives at worldcup.betminer.co.uk. It takes all 48 teams, plays every group fixture, every knockout tie, extra time and penalties where the script demands them, crowns a champion, then wipes the board and does it again. Ten thousand complete tournaments, each one sampled from probabilities rather than chosen by anyone with a favourite shirt.
The result is not a single prediction. It is ten thousand of them layered together, which gives you a probability for how far each team is likely to go. Here is what came back.
Nobody clears 13 percent
Across the 10,000 runs, Spain finished as champions more often than any other side, lifting the trophy in 12.9 percent of them. That makes them the model's favourite. It also makes them the least convincing favourite a World Cup has produced in a long time.
That 12.9 percent is the whole point of the exercise, so it is worth sitting with for a second.
The tournament has grown to 48 teams. Where there used to be eight groups there are now twelve. There is a Round of 32 bolted on in front of the Round of 16, and the match calendar has swollen from 64 games to 104. Every additional knockout round is one more flip of the coin that the best team in the field has to win. Stack those flips on top of each other all the way through the bracket and even the side the algorithm rates highest only ends up champion in roughly one tournament out of eight.
Turn that around and it reads more starkly. In 87 percent of the tournaments we simulated, the winner was someone other than Spain. That is not the model being timid. That is what a 48-team knockout does to certainty.
Behind Spain, France came second on 10.1 percent and Argentina third on 8.4 percent. Then the floor drops away. Morocco are fourth on 5.0 percent, Japan fifth on 4.9 percent, England sixth on 3.6 percent, and once you are past the leading seeds the numbers shrink into fractions of a percent.
What is actually doing the work
The full write-up sits on the methodology page, but the shape of it is straightforward enough to explain here.
Underneath everything is international football Elo. It is the approach FiveThirtyEight built its ratings on for years, and it is close to what FIFA rebuilt its own ranking system around in 2018. Each team holds a rating, every result nudges that rating depending on the scoreline and the quality of the opponent, and the gap between any two ratings hands you a baseline probability for a match between them. It is well established, it holds up in the academic work, and it is about as unglamorous as a reliable foundation usually is.
On top of that sits the Betminer signal layer. Form trend, the spread of winning and losing margins, defensive reliability, how a side has looked recently against tactically similar opponents. These are the same signals the Betminer Algorithm has been sharpening across seasons of week-to-week football prediction, and the same ones that power the daily selections you will find through the network.
The hybrid sits the two together. Elo gives the anchor probability, and the recent signals tug it one way or the other. A structurally strong team going through a flat patch gets marked down. A team Elo has overlooked because its recent fixtures were soft gets marked up. The adjustment is capped, so a couple of flattering friendly results cannot send the whole thing haywire, and what comes out for each match is a range of outcomes rather than one tidy number.
Then the simulation runs. Each match is drawn from its own distribution, the group tables settle, the knockout draw unfolds, and a champion emerges. Do that ten thousand times and you have ten thousand separate World Cups. Whichever team wins the most of them is the model's pick.
The bracket the model expects
The single most likely route through the draw ends with Spain beating Argentina in the final. France go out to Spain in one semi-final, Morocco go out to Argentina in the other.
That Morocco run is one of the model's more eye-catching calls. It has them topping a difficult Group C ahead of Brazil and Scotland, then handed a winnable Round of 16 tie against England, then winning a quarter-final the model expects to fall their way.
Which does, yes, mean the model knocks England out at the quarter-final stage against Morocco. We will deal with that in a moment.
Argentina's path is the other one that lingers. The holders are the third most likely champions in the whole field, yet the model walks them all the way to the final and then has them lose it. The gap from second to third is statistically tiny, 10.1 percent against 8.4 percent, but it lands on a clean storyline anyway: Messi reaching one more final, and one more final slipping away.
The glaring absence is Brazil. The five-time winners came out ninth on 1.8 percent, with the bracket sending them home in the Round of 16 at the hands of Spain.
Where the betting value sits
Here is the part that matters if you back football with real money rather than just settle the argument on the way home.
Two genuine overlays fall out of the model. A third looks like bait.
Japan to reach the quarter-finals at 7/2 is the one we would put first. The model makes them the most likely winners of Group F, ahead of the Netherlands, sends them through a kind R32 tie, and into the Round of 16 in a clear majority of runs. From there it is one more knockout game against the sort of opponent the bracket tends to throw up. At 7/2 the bookmakers are implying a 22 percent chance. With Japan at 4.9 percent to win the tournament outright, they are reaching the last eight in comfortably more than a quarter of all simulations. The price on the board is short of where our line sits.
Morocco at 150/1 outright is the second, for a small stake. They are not the base-case winners, but they reach the semi-finals in 12.3 percent of runs and the final in 6.9 percent. A 150/1 price on a team the model rates at roughly one-in-twenty to go all the way is the sort of overlay that only exists because the market is still pricing African sides off old reputation rather than current form.
Brazil at 9/1 is the one to walk past. That price implies a market view of around 10 percent. The model has them at 1.8 percent. That is a fivefold gap. Either we have badly misjudged the current Brazil, or the entire market is still pricing the yellow shirt on history. We would back the second reading. If anything, the sharper value may be in a Brazil early-exit market rather than anywhere near backing them to win.
Read the numbers properly
A few things need saying plainly before anyone gets carried away.
This is a probabilistic model. It is built on the best data we could assemble, checked against past tournament results, and tested against the recent form of every side in the field. It is not a crystal ball. The reason for running ten thousand simulations instead of naming one winner is precisely to show you how uncertain the picture is, and what that uncertainty looks like.
Spain at 12.9 percent means Spain at 12.9 percent. It does not mean Spain to win. Read it as: play this tournament a hundred times and Spain come out on top in about thirteen of them. That is a very different claim from saying Spain are going to win the World Cup.
And none of this is betting advice. A model output is one input among several. Your bankroll, the available prices, your own reading of injuries and form between now and June, all of that goes on top of what the numbers here are telling you.
If you want the full picture, the model carries probabilities for all 48 teams, a predicted bracket showing the consensus route to the final, and a breakdown of the likely finishing order in every group. It is worth a proper look before the tournament starts. If you want to get all the best tips daily through the World Cup from this model then join our VIP, which is powered by this model: footballtipster.ai/world-cup.
For now the summary is simple. Spain are the model's favourites by a nose, the field is far more open than the prices suggest, Japan look like the smartest tournament-stage bet available, and the odds are heavily against the five-time champions being the team that lifts the trophy in New Jersey on 19 July. Probably.