My 2026 World Cup tournament simulations and projections are now live!
The widget below will show you my full 2026 World Cup projections, so you can see how far I project each team to advance.
Check back daily for updates, because as the tournament goes along, I will be updating this regularly. The widget below will have a timestamp, so you can see when this was most recently updated.
World Cup Projections
The 2026 World Cup Model Explanation
A few things to note about these projections.
This model was trained on international matches from just after the 1994 World Cup until present, which is a fitting era since a few major things happened.
First, FIFA moved to three points for a win instead of two, and second, the 1998 World Cup was the first to expand to 32 teams.
Every game is weighed by its importance (friendly, minor tournament, qualifier, major tournament, World Cup, etc.) and its recency, with two different decay rates. One is more ELO-based, better reflecting current form, while the other estimates each team's attack and defense coefficients using Maximum Likelihood Estimation from its past results with a longer decay weight. The two versions are then blended with their own appropriate weights, and then adjustments to those resulting values are made for current rosters.
To test this model, it's fit out of sample on all major international tournaments from 2010 to present, so the model only "sees" data prior to the tournament and then tests on that tournament. Doing it this way led to some important findings.
Notably, when simulating each game, this model uses something called diagonal-inflated bivariate poisson to create score-table distributions for each game.
While it's commonly been shown that 0-0 and 1-1 scorelines are inflated, that often is for international friendlies or club league plays. Empirical testing of major tournaments since 2010 has shown a bit of a deflation of these scorelines, and the game-by-game projections will reflect that.
Other score distribution methods, such as negative binomial were tested, but rejected after rigorous testing showed no out-of-sample improvement in major international tournaments.
This model also uses some novelties around inter-confederation vs. intra-confederation play (think of it like NFL divisional games vs. out-of-division games) as well as how it treats group stage vs. knockout round desperation (the "bivariate" in the model's name).
Additionally, for about the last decade, expected goals (xG) data has been available for many international games and all major tournaments. This model uses an emperically optimal blend of actual goals and xG for games where xG data is available.
Home-field advantage isn't applied equally. Mexico, at altitude (especially in Estadio Azteca), gets a larger home-field advantage boost than the USA does in L.A. or Seattle, or Canada does in Vancouver.
This model also uses a "hot" updating method where each team's attack and defense coefficients are recalculated after each game played. That way, if Cape Verde beats Spain, their coefficients would improve, and Spain's would decline. That allows for a more natural variation in the tournament rather than a "cold" simulation. where each team's coefficients are frozen before the tournament and used for every game played.
Also, there are 495 combinations of third-place results. This model accounts for that, so it accurately reflects which teams play each other in the Round of 32 and beyond.
2026 World Cup Model Projections and Simulation Results
After 25,000 simulations of the World Cup, here are the 2026 World Cup probabilities for each team. Below the widget is some interesting discussion, as some of this model's results might slightly defy conventional wisdom.
(Note: you can change the color scale in the widget below.)
Pre-Tournament Discussion
On the simulation results themselves, notably contrary to the market and popular wisdom, Brazil is second in tournament win probability by this model. That's not because they are the second-best team. Instead, this model has them around fourth or fifth best, but their path is a bit easier, especially in the group stage where they are 67.3% to win their group compared to France's 61.3% in a much tougher group that includes Norway and Senegal.
Another factor lowering France's odds is their strong chance of facing Germany, another UEFA superpower, in the Round of 16 if each favorite advances. By comparison, Brazil would get the runner-up from either Group E or I, instead of facing a group winner, if the chalk script plays out.
More discussion will be added below with interesting results as the tournament unfolds.
