How the simulation works

Every result on this site is produced by one model. It is deliberately simple — a rating for each team, a scoring rate for each side of a match, and dice — and it is tuned against real international results rather than against what looks exciting. This page sets out what it does, the numbers it uses and what it leaves out.

1. How strong is each team?

Each of the 211 nations has three numbers.

Rating. This is the nation's points total in the FIFA/Coca-Cola Men's World Ranking, which since 2018 has itself been an Elo-style system: teams gain points for results that beat expectation and lose them for results that fall short, with bigger matches counting for more. The simulator currently uses the ranking published on 7 October 2026 and is updated when FIFA publishes a new one. A game you have already started keeps the ranking it began with, so a refresh never rearranges a tournament in progress.

Attack and defence, each from 1 to 100. These come from a nation's last 30 senior internationals in an open dataset of results. Goals scored and conceded are compared with what the two teams' ratings would have predicted for that fixture, weighted by how recent the match was and by its importance — a World Cup match counts for more than a friendly. The result is then pulled back towards the team's overall rating, more firmly for lower-ranked teams, so that a handful of big wins against weak opponents cannot turn a minor nation into a scoring machine.

You can change all three numbers for any nation in the simulator's team editor.

2. From ratings to a scoreline

For each match the model works out how many goals each side should expect, then draws the actual score at random around those expectations.

1. Win expectancy. The difference between the two ratings is turned into an expected share of the result with the standard Elo curve, 1 / (1 + 10^−gap/400). Equal teams get 50% each; a 200-point gap gives the stronger side about 76%.
2. Scoring rates. Both sides start from a base of 1.25 goals. That is moved up for the favourite and down for the underdog in proportion to the win expectancy, multiplied by the team's own attack and by the weakness of the opposing defence, and finally scaled by 0.93 so that total goals match the real game. Attack and defence act as multipliers between 0.85 and 1.15.
3. The score. Each side's goals are drawn from a Poisson distribution around its scoring rate — the usual model for goals, in which a team expecting 1.5 scores exactly none about 22% of the time and four or more about 7% of the time.

Two refinements sit on top. When the rating gap is over 400 points the weaker side's scoring rate is cut by a further 40%, because real mismatches are more one-sided than a smooth curve suggests. And 7% of matches that would finish with one goal between the sides end level instead: independent scoring is known to produce too few draws, since a team that is a goal up tends to protect the lead and a team that is a goal down pushes. That adjustment adds about 2.9 percentage points of draws between evenly matched sides. It is never applied to a knockout tie.

Worked examples

These five fixtures have been run through the engine exactly as the simulator would, with today's ratings. "Expected goals" are the two scoring rates; the percentages are the chance of each result over ninety minutes.

3. Home advantage

A team playing in its own country is treated as 50 rating points stronger for that match. That applies to the home side in every qualifier, to Morocco, Portugal and Spain whenever they play in their own country during the finals, and to Argentina, Paraguay and Uruguay in their centenary opening match. Matches on neutral ground — most of the World Cup, and the whole inter-confederation play-off — carry no advantage for either side.

The size was set from real qualifiers. In home-and-away qualifying since 2015, taking only pairs of teams that hosted each other so that the difference in strength cancels out, the home side won 48% of matches, drew 22% and lost 30% — an edge of about half a goal a game. Fifty points is the setting that comes closest to that.

What the model does not know is why a venue is hard. Altitude in La Paz, heat, long-haul travel and artificial pitches are all folded into the same flat bonus, which is why Bolivia at home in the table above are still clear underdogs against Brazil when the real fixture is famously awkward.

4. Is it realistic? The calibration

The model's settings were chosen by comparing whole simulated competitions with real ones.

Home and away wins are not shown for the World Cup because nearly all of its matches are on neutral ground. The model still gives home sides slightly less than they get in reality; pushing the bonus further would have meant leaving the range the match editor exposes, so it was left as the closest honest setting.

5. Knockout ties and penalties

A knockout match is played like any other. If it finishes level it goes to a shoot-out; extra time is not played out as a separate period. The shoot-out is close to a coin toss, leaning slightly towards the stronger side: its chance is 50% plus one percentage point for every 40 points of rating gap, so a 200-point favourite wins 55% of shoot-outs. The score of the shoot-out shown in the simulator is generated to fit the winner.

6. Tables and tie-breakers

Group tables are ordered by points. Teams level on points are separated by the results between them — points, then goal difference, then goals scored in those matches — and that is applied again to any teams still level. Only then do overall goal difference and overall goals scored come in, and as a last resort the FIFA ranking.

In the 48-team format the eight best third-placed teams are ranked across all twelve groups on points, goal difference, goals scored and then FIFA ranking. Their opponents in the round of 32 are fixed by which groups they come from, using the pairing table of the 2026 tournament.

7. Draws and seeding

Qualifying groups are seeded by FIFA ranking, so the strongest nations in a confederation are spread across the groups rather than drawn together.

For the finals the qualified teams are split into four pots by ranking. Morocco, Portugal and Spain are placed in pot one and into the first three groups. The other pots are then drawn at random with one constraint: no group may contain two teams from the same confederation, except Europe, which may have two. Each draw comes from a seed, so the same game always shows the same groups however many times you reopen it.

8. Fatigue and match conditions

In the finals a team's attack and defence are reduced a little as the group stage goes on — by 3% in its first match, 4.5% in its second and 6% in its third — to stand in for accumulated fatigue. Qualifiers and knockout matches use the lowest setting.

The match editor lets you override the conditions for a single game: weather, crowd support, morale, fatigue and whether each side is missing its star player. Automatic simulations use neutral conditions throughout, apart from home advantage.

9. Where the odds come from

The percentages quoted across this site — a nation's chance of qualifying, of reaching the quarter-finals, of winning — are counted, not calculated. The entire cycle was played 20,000 times from first qualifier to final: 639 qualifying matches, a fresh draw and 104 finals matches each time, about 15 million matches in all. A nation that "qualifies 31%" reached the finals in 31% of those cycles. With that many runs a figure near 50% is accurate to within about one percentage point; very small figures are rougher.

The 64-team scenario was played 10,000 times through qualifying only.

10. What the model leaves out

• Players. There are no squads, injuries, suspensions or managers. A nation is its rating.
• Time. Ratings are today's. The real qualifiers run from 2027 to 2029 and the finals are in 2030; some of today's strongest sides will be weaker by then and others will have improved.
• Style and match-ups. Every team plays the same way, scaled by its numbers.
• Venue specifics. Home advantage is one flat bonus, as described above.
• The real draw and the real formats. Groups are drawn by the simulator, and where a confederation has not yet published its 2030 format the simulator uses the shape of the previous cycle. See qualifying explained for what is official.

So treat the numbers as a description of how this model behaves, and as a way of seeing how hard each route is — not as a prediction of what will happen in four years.