The 2026 World Cup is using a new format: 48 teams, 12 groups of 4, with the top 2 from each group plus the 8 best third-placed teams advancing to a brand new Round of 32. That sounds like a small wrinkle. It's actually the biggest structural change to the tournament since 1998 and it's going to produce the most upset-prone knockout round we've ever seen.
We ran 1,000 simulations with our Poisson + confederation model. About 41% of sims had four or more outright upsets in the R32. 18% had six or more. For comparison, in the old 16-team Round of 16 format from 2010 to 2022, the historical average was 1.5 to 2 upsets per round. The 2026 R32 will roughly triple that.
Here's why, in three structural reasons that compound.
1. The qualifying bar drops
In the old format, a team finishing third in their group went home. They had to win the group or finish second to advance. That's it. So group stage was a serious filter.
In 2026, a team can finish THIRD in their group with as few as 4 points (one win, one draw, one loss) and still advance. Because there are 12 groups and only 8 of the 12 third-place finishers go through, FIFA picks the best 8 by points-then-goal-difference. That means a third-placer from a strong group with 4 points can knock out a third-placer from a weak group with 3 points.
Net effect: teams that wouldn't have advanced under the old rules are advancing. They're entering the knockouts a half-tier weaker than the old R16 field. The R32 has a wider talent spread than any knockout round we've seen.
2. Path mismatches are baked in
The 32-team bracket pairs seeds 1 through 32 (1v32, 16v17, 8v25, etc). FIFA reserves seeds 25 to 32 for the best third-placers. So if you're the top seed (probably Argentina or France in our model), your R32 opponent is by definition a third-place finisher from somewhere.
But here's the catch: not all third-place finishers are equal. A 3rd-placer from a brutal group like F (Netherlands, Japan, Tunisia, Sweden) could easily be Sweden, who'd be a stronger team than several Group G winners. Yet Sweden gets a 3rd-place seed and lands against a top-8 seed, while a weaker Group G winner cruises through their R32 as a top-12 seed.
The old format avoided this because group winners always met group runners-up. Now winners meet third-placers, and the strength gap is unpredictable.
3. The rest disadvantage doesn't apply
The 2026 R32 plays from June 28 to July 2. Group winners get an extra day or two of rest depending on which group, but third-placers are mostly fresh too because they've also been playing since June 11. Rest differential is at most one day.
In tournament football, one day of rest is barely a factor. So the slight rest advantage that group winners enjoyed in old formats has functionally disappeared. The R32 is a level playing field on fitness, which means skill differential is the only filter, which means upsets become statistically more likely.
What the simulations actually show
Across 1,000 sims of the full tournament, here's what we see in the R32:
- •Probability of 4+ outright upsets (lower-seed wins): 41%
- •Probability of 6+ outright upsets: 18%
- •Probability of 8+ outright upsets: 5%
- •Median number of upsets: 3 (vs historical 1.5)
- •Probability that at least one top-8 seed loses in R32: 82%
That last number is the one to dwell on. In 82% of our simulations, at least one of the top 8 teams (Argentina, France, Brazil, Spain, England, Germany, Netherlands, Portugal) crashes out before the Round of 16. Historically that figure was around 25%.
Which teams are dangerous as third-placers
Based on group projections, the third-place teams most likely to cause R32 upsets are:
- •Sweden (likely 3rd in Group F if Netherlands and Japan finish 1-2)
- •Norway (could finish 3rd in their group depending on draw)
- •Croatia (always dangerous in knockouts, our model rates them at 1.9% to win the whole thing)
- •Senegal (CAF teams are systematically underrated in resale markets)
- •Mexico (if they finish 3rd, which our model gives a 22% chance)
Any of these as your team's R32 opponent is a worse draw than the seed numbers suggest.
What it means for your bracket
If you're filling out a bracket: don't auto-advance the top-seeded team in their R32 match. The model gives the top 8 seeds an average of 78% to advance through R32, which means roughly 1.7 of them will lose. Pick at least one upset.
If you're picking your champion: the chaos affects the path more than the destination. Our model still gives Argentina (14.2%), France (11.8%), and Brazil (10.4%) the highest title odds. But the route through is more volatile than ever, so don't over-weight a smooth-looking bracket on paper.
You can run your own simulation, swap in any group-stage results you expect, and see how the bracket reshapes at myworldcupguide.com/predictor. The R32 pairings update in real time as you change group standings.
Bottom line
The format change isn't cosmetic. Adding 16 teams and a Round of 32 is the most consequential structural change to the World Cup in 28 years, and the math works against the favorites. Expect at least one of the big eight to lose before the Round of 16. Plan your viewing accordingly.