The 2026 World Cup advances eight of the twelve third-placed teams into the Round of 32. That much is widely known. What is far less understood is how those eight teams get assigned their opponents, because the assignment is not intuitive, it is not fully fixed until the group stage ends, and in two corners of the bracket it is completely locked before a single group game is played.
This is a closer look at how the third-place route actually works in the simulator, built on the same 100,000-simulation dataset behind the team profiles. The earlier write-up on the tournament as a path problem covered why the bracket is hard to draw at all. This one goes one level deeper, into the routing table that decides who plays whom.
Why the third-place route is different
Finishing first or second in a group leads to a known bracket slot. The path may be tough, but it is fixed by position. Third place is not like that.
Twelve groups each produce a third-placed team, but only eight advance. The eight that survive are selected by ranking all twelve against each other on points, then goal difference, then goals scored. None of those twelve teams played each other, so this is a cross-group comparison: a team can finish third with the same record in two different simulations and qualify in one but not the other, depending entirely on results in groups on the other side of the bracket.
Once the eight qualifiers are known, the official regulations assign each one to a group winner using a fixed lookup table. There are 495 possible combinations of which eight of twelve thirds advance, and the table has an entry for every one. The simulator reproduces this table in full.
A note on that 495 figure, because it is easy to overstate. It is the number of distinct ways to choose eight qualifying thirds from twelve groups, and therefore the number of routing scenarios the table must cover. It is not the number of ways the tournament can play out, which is vastly larger. The 495 describes the routing problem specifically, nothing more.
The cutoff is usually three points, but three points is not a guarantee
Across the simulation runs, the qualification cutoff for the eighth and final third-place spot most often lands at three points. In the large majority of runs, a third-placed team on three points is in contention.
But being on three points does not settle it. In the runs where the cutoff sits exactly at three points, only around two thirds of the teams on that mark actually advance. The rest are edged out on goal difference and goals scored. This is a model output rather than a fixed rule, and the exact share shifts with the seed and calibration, but the direction is clear: at the margin, every goal in every group game can decide whether a third-placed team survives, even goals scored in a group that team is not part of.
For a fan, this is the strange part of the format. A team can finish its own group, sit on three points, and then spend two days watching unrelated matches to learn whether it advanced.
Annex C: the part that is actually locked
Here is where the routing turns from probabilistic to fixed.
Most third-place assignments vary with the combination of qualifiers. A third-placed team from Group C, for example, can be routed to the winner of Group E, Group A, or Group I depending on which other thirds came through. In the simulation, that comes out as roughly Germany most of the time, Mexico a good share of the time, and France occasionally. None of those is guaranteed, because the assignment depends on the wider picture.
Groups K and L are the exception. In every single routing scenario where both groups send a third-placed team through, the table sends them across to each other. The Group K third is assigned to the Group L winner. The Group L third is assigned to the Group K winner. Checked across all the relevant combinations in the table, there are zero exceptions.
It is worth being precise about what is locked here. The table fixes the slot, not the identity of the opponent. The Group K third is sent to whoever wins Group L, and the Group L third is sent to whoever wins Group K. In the current 100,000-simulation snapshot those winners are most often England and Portugal, which is why a third-placed DR Congo, Uzbekistan, Ghana or Panama tends to pick up a very sharp-looking Round of 32 draw in the profiles. But if a different team were to top Group K or Group L, the crossover would still hold and simply point at that team instead. The locked part is the routing, not the name.
This is not a quirk of the model. It is a deliberate structural choice in the official bracket, faithfully reproduced. For West and Central African football audiences in particular, it makes for a sharp framing: the simulator can tell Ghana its chance of qualifying as a third, but the routing table has already decided which group’s winner is waiting if they do.
Why a single opponent list can mislead
There is a practical consequence of all this for reading the team profiles, and it is worth being explicit about.
Each profile shows a general opponent pool for each round, aggregating every simulated path. For a team whose route runs through several finish positions, that pool blends together opponent distributions that never actually occur together.
Scotland is the clearest example, and it is the one that came up directly when the profiles went public. Scotland’s general Round of 32 pool is led by the Netherlands, with Japan, Germany and Sweden close behind. Read as a single list, it looks like one coherent set of likely opponents. It is not. The Netherlands appears almost entirely through the first and second place paths, where Scotland is routed against Group F. Germany appears almost entirely through the third place path, where Scotland is routed against the Group E winner. A third-placed Scotland never plays the Netherlands. The two names sit in the same list but belong to different worlds.
This is why the profiles separate the opponent pools by finish position. A general pool answers “who might this team face overall,” which is genuinely useful, but it cannot answer “who will this team face if it finishes third,” and treating it as if it can produces exactly the confusion above. The path-by-finish view exists to keep those three routes apart.
How to read these numbers
Everything here comes from a 100,000-simulation Monte Carlo run with a specific model calibration. The figures are frequencies across those runs, not forecasts and not betting guidance. A statement like “the Group C third faces Germany most of the time” means that across the simulated runs where that path occurred, Germany was the most common opponent. It is not a prediction that any particular team will finish third, and the slot percentages for most groups are probabilities, not certainties.
The one assignment that is genuinely fixed is the K/L crossover, because it is fixed in the routing table itself rather than emerging from the simulation. Everything else is a distribution.
The model also makes deliberate choices. Upset behaviour is constrained rather than wide open, team strength is driven by squad ratings, and the third-place tiebreaker in the model stops at goals scored, where the full FIFA procedure includes further steps. None of this changes the structural points above, but it is the honest backdrop to the numbers.
See the routing for yourself
The team profiles show the full path-by-finish breakdown, including the separate third-place opponent pool for every team that has a meaningful third-place route.