The 2026 FIFA World Cup is one of the most watched events of the international sports calendar, and fans from across the globe will be trying to predict how far their team will go.
I’m a data scientist and in an attempt to forecast the eventual tournament winner, semi-finalists and teams’ chances of progressing through the group stages, I built a model to predict how the World Cup may unfold.
Here’s how I did it and what my model predicted.
Lessons from recent history
For this World Cup, the traditional 32-team tournament structure (eight groups of four) has been expanded to a bulging 48 teams (12 groups of four), with new progression rules, an extra knock-out round and a rise in total matches from 64 to 104.
The changes were designed by FIFA primarily to increase global participation, maximise revenue through more matches and boost the popularity of soccer in new markets.
Read more:
Curaçao and Cabo Verde are into the World Cup. What impact can these ‘minnow nations’ make?
In trying to predict the 2026 event, what can recent history teach us?
Looking back to the seven 32-team tournaments since 1998, the 28 semi-final spots have been dominated by six nations who reached that stage more than once: Argentina (2), the Netherlands (3), Brazil (3), Croatia (3), France (4) and Germany (4).
If we include previous tournament winners (England, Italy and Spain), 78.6% of the modern semi-finalists have come from nine nations.
Further, all 14 finalists were from this group – the last finalist from outside these nine came in 1962 (Czechoslovakia); the last winner was in 1950 (Uruguay).
This is an amazing degree of dominance given the number of international teams playing the game – official FIFA rankings currently list 211 nations.
More teams at the 2026 event, though, means it is harder to accurately assess the likelihood of tournament results.
For this, we need analytics, and I’ve undertaken a simulation study designed to calculate the progression chances of all 48 teams in the field.
While the obvious outcome of such a study is to assess who the likely winners are, we can also gain insight from how the new format spreads these chances across the teams and how it affects the chances of the top sides raising the trophy.
What did it predict?
Each team’s chances of reaching each round, based on one million simulations, are shown in the below table.
It predicts Australia is a 67.1% chance of getting out of their group, a 31.3% chance of getting past their first knock-out match, but is just a 1.0% chance of making the final and 0.3% chance of winning.
Canada’s chances are quite similar: a 78.9% chance of making out of their group (thanks to being a host nation), a 37.9% chance of getting past their first knock-out but just a 1.0% chance of making the final and 0.3% chance of winning.
New Zealand, on the other hand, has basically no chance of winning and only a 19.5% chance of making it out of their group.
Lastly, while England has the fourth highest overall chance of winning, it is notably lower than the other three favourites. This is at least in part due to their recent drop in rating after a loss to Japan in March.
The only teams with more than 10% chance of winning the trophy are Spain (15.8%), France (15.6%), Argentina (15.3%) and England (11.0%) – all members of the “group of nine” and the current top four rated sides.
But the estimated proportion of semi-final spots taken by these nine nations is 54.2% – notably lower than the historical 78.6%.
Further, the estimated proportion of finalists to come from these nine nations is 63.6%, while there is a 72.6% chance the champion comes from this group, both down from the historical 100% values. Of course, this is partly due to Italy’s failure to make the World Cup.
So, FIFA’s new format does reduce the chances of the historically strong nations progressing far into the tournament but not as much as they may have hoped.
Had FIFA increased the size of the groups to six teams, instead of increasing the number of groups, the new format would have done more to spread the chances – but doing so would have required at least 136 matches.
I would like to thank Dr Chris Bilson and Noah Stern for their help in producing this article. In addition, I would like to thank a very careful reviewer for important comments.
