We ran the 2026 season a thousand times. Here is what a season looks like.

The question

Every season simulator, including ours, produces a champion when you press the button. The trouble with one champion is that it tells you almost nothing. Was that the likely result, or the one-in-twenty? You cannot know from a single run, in the same way that you cannot tell whether a coin is fair from one toss.

So we ran it a thousand times. A thousand complete seasons of the 2026 grid, each one a full year of three-part qualifying, tyre strategy, safety cars, retirements and weather, scored under the real points system. Twenty-four thousand Grands Prix. What follows is what the model believes about this grid once the noise has averaged out.

How this was measured

The engine is the same one behind every game on this site, and it is run headless: no player, no team mode, nobody spending a development budget. Every team's form still swings from weekend to weekend, as it does in the game, so this is not a fixed pecking order being replayed — it is the same starting grid rolled through a thousand different years.

Two things worth saying before any number. The driver and car ratings underneath this are our model, not official data. And they were re-calibrated against the first thirteen rounds of the real 2026 season before this run — the constructors' order and every team-mate split — because a pre-season guess that had McLaren on top was, by September, simply describing a different year. Where the results below look confident, that is the model being confident, and we point at two places where we think it is too confident.

Russell takes the poles. Antonelli takes the titles.

Across a thousand seasons, Russell qualifies on pole more often than anyone: 7.1 times a year. Antonelli manages 6.8. And yet Antonelli takes the title in 741 of the thousand seasons, Russell in 220. The same car, a near-identical Saturday, and a three-to-one split on Sunday.

The gap is in the columns nobody puts on a poster. Antonelli converts his Saturdays: 7.0 wins a season to Russell's 4.7, 16.2 podiums to 14.4, and one fewer retirement every ten years. Over 24 rounds that compounds into a 33-point average gap, and the median title margin in this set is 36. Qualifying decides where you start; the model's racecraft and consistency ratings decide where you finish, and that is where the two Mercedes drivers separate.

The two Ferraris are the opposite story: Hamilton and Leclerc finish the year on 307 points each, Leclerc with more poles (4.7 to 1.4), Hamilton with fractionally more podiums, and the team-mate battle goes 512–488 over a thousand seasons. That is as close to a dead heat as a model can produce.

Below the top two, the field is in layers

The standings table above is the median finishing position across all thousand seasons, and what it shows is not a smooth slope. It is steps. The two Mercedes drivers are a group of their own: both finish in the top two in most years, and the same team fills first and second in 74% of seasons. Then a sixty-point gap to the Ferraris. Then Verstappen, Norris and Piastri, sharing the next three places between them. Then a two-hundred-point drop to the midfield, where the entire rest of the grid fights over what is left.

Verstappen is the odd one out. In the fourth-best car on the grid he averages 282 points, which is more than either McLaren: Norris on 277, Piastri on 242. He wins three races a year and finishes on the podium eight times in a car whose second driver scores 70. That is the model's driver ratings doing what they are for.

Mercedes wins the constructors' championship 968 times

Not most of the time. Nearly all of it. The drivers' title is a genuine contest between two people in the same car; the constructors' title is barely a contest, because it counts both cars, and Mercedes is the only team with two drivers who finish in the top two more often than not. Ferrari takes the other 32.

That is also why the drivers' champion comes from a team other than the constructors' champion in only 4% of seasons. Last year's version of this grid produced that result four times in ten; a dominant car with two strong drivers closes the door on it.

Four in ten championships go to the last race

The number we were most curious about, and the one that survived the recalibration almost untouched. In 402 of the thousand seasons the drivers' title was still mathematically open going into the final round. Another 355 were settled at the penultimate race, 169 at the one before that. Only nine seasons in a thousand were decided with four rounds to spare, and no title was ever clinched before round 20.

The margins match. The median gap between champion and runner-up is 36 points, a little over one race win. In 15% of seasons it is under ten points, which is the difference between a second and a fourth somewhere in March. In 17% of all seasons, the champion did not win the most races.

The back of the grid

Eleven teams, and the bottom seven average 178 points between them across a whole season — less than Verstappen scores on his own. A point at this end of the grid is a rare event, which is why the tail of the table is where the model is least predictable: Alpine's median finish is fifth, but it finishes anywhere from fifth to tenth, and every team behind it has finished as high as fifth and as low as eleventh within the same thousand years.

The number that stood out: Cadillac, the new entry, scores nothing at all in 7% of seasons, and Stroll's median finish is 21st with a season best of 16th. Nobody else ever ends a year on zero. In the real sport a pointless season is an existential event; in the model it is one Cadillac year in fourteen, which is a rating the rest of the real season will test.

Where we think the model is too sure

A model that agrees with itself a thousand times is not the same as a model that is right, and two results here are the ratings speaking rather than anything we would bet on.

• Norris beats Piastri 84% of the time. The real split after thirteen rounds is 171 points to 116 — Norris ahead, clearly, but nearer three-to-two than five-to-one. The model has read that gap as a pace-and-consistency difference and then compounded it over 24 rounds. It is the first rating we would revisit.
• A Mercedes one-two in 74% of seasons. Antonelli and Russell are clear of the field in the real standings too, but a car that good with two drivers that consistent leaves almost nothing for Ferrari, and one upgrade in Maranello would change that faster than a rating file does. Treat the constructors' number as an upper bound.

What the model is not doing is what the simplest season predictors do: ranking drivers by a single number and adding noise. Every one of these standings is the sum of races that were actually run, lap by lap, with track position kept honest — a quick car that qualifies badly still has to find a way past. That is why Russell can take the most poles and lose the title, and why Verstappen can outscore both McLarens from the fourth-best car, which are results a single-number model cannot produce.

Run your own

Every number on this page came out of the F1 championship simulator, and you can run a season in a second, or a round at a time to watch the title swing. For the real title fight — the actual standings, who is still alive and what Antonelli needs to clinch — use the championship calculator. If you want the odds for a single Grand Prix rather than a year, the race predictor does the same thing per circuit, and if you want to know why Monaco decides so little from the back of the grid, we measured that too.

Questions

Who wins the 2026 F1 championship in the simulation?

Kimi Antonelli, in 74.1% of a thousand simulated seasons. George Russell wins 22.0%, Lewis Hamilton 1.8%, Charles Leclerc 1.0%, Max Verstappen 0.7% and Lando Norris 0.4%. Mercedes won the constructors' championship in 968 of the thousand seasons and Ferrari in the other 32.

Why does Russell take more poles but fewer titles?

He averages 7.1 poles a season to Antonelli's 6.8, but converts them into 4.7 wins against Antonelli's 7.0, with fewer podiums (14.4 to 16.2). Over 24 rounds the Sundays outweigh the Saturdays, and the median title margin is only 36 points.

How often does the F1 title go to the last race?

In this model, 40% of seasons are still open at the final round, and 76% are decided at one of the last two races. Fewer than 1% are settled with four rounds to spare, and no title was ever clinched before round 20 of 24.

Is this a prediction of the real 2026 season?

No. It is a measurement of our model of the 2026 grid: our own driver and car ratings, calibrated to the first thirteen real rounds and then run through our race engine a thousand times from the start of a season. The ratings are not official data, and the article points at two results we think the model is too confident about. For the real standings and what the leader needs, use the championship calculator.

How long does it take to simulate a whole F1 season?

In the browser, about a second for all 24 rounds, or one round at a time if you want to follow the title fight. Headless, the thousand seasons behind this article took just over a minute.

More on this site

• F1 Career Simulator — one driver, karting to Formula 1.
• F1 Simulator Game — a single Grand Prix, lap by lap.
• F1 Championship Simulator — a full 24-round season, or run a team.
• How the simulation works — the model, in detail.