How many races does it take for the best driver to win? 2,700 seasons measured
F1 Legend Sim · Staff · 18 August 2026 · 8 min read
Every time a championship goes to the final race, the same argument comes back: is a short calendar a lottery, and does a long one deliver justice? It sounds reasonable, and nobody has ever checked, because 2007 was only run once. This analysis ran it a hundred and fifty times over five races, another hundred and fifty over ten, over sixteen and over twenty-two. Then we did the same with 2010, 2021 and 2024. The answer is not the obvious one: the number of races barely matters — what matters is how close the cars are.
What exactly was measured
We take a real grid from the database — the twenty drivers of that year, with their attributes and with the performance and reliability of their cars — and run the whole championship. Then we run it again with a different seed, and again, up to a hundred and fifty times. The only thing that changes between runs is chance: the order of the circuits, the rain, the retirements, the overtakes that come off and the ones that do not.
That is repeated with calendars of 5, 10, 16 and 22 races. Sixteen cells per season, 2,400 complete championships in total, plus 300 for an experiment we get to at the end. The engine is the one that drives the rest of the site: weighted formulas over attributes, car, weather, reliability and strategy, with its parameters calibrated against real Formula 1 results. If you are curious about how it is built, we explain it in how a Formula 1 simulator works.
One limitation, stated up front rather than buried in a footnote: the tool that runs full seasons always puts a control driver in one seat — in these experiments, the second Ferrari — with average attributes. It does not change who the favourite is in any of the four seasons, but it is a limitation that has to be declared — and that control driver ends up producing one of the figures in this analysis.
The table: how often the favourite wins
“Favourite” here is not an opinion: it is the driver who wins most championships over the long calendar, i.e. the one the grid says is best when you give him time to prove it. The question is how often that same driver wins depending on how long the year is.
| Season | Favourite | 5 races | 10 | 16 | 22 | Margin |
|---|---|---|---|---|---|---|
| 2007 | Hamilton | 49% | 66% | 67% | 67% | 7.9% |
| 2010 | Vettel | 91% | 96% | 99% | 100% | 28.9% |
| 2021 | Verstappen | 63% | 71% | 77% | 88% | 9.8% |
| 2024 | Verstappen | 98% | 99% | 100% | 100% | 33.0% |
The last column explains everything else: the margin is the champion’s average lead over the runner-up, as a percentage of his own points, over the 22-race championship.
In a lopsided year, five races are already enough
Take 2024. With a five-race calendar — the 1950 season, roughly — Verstappen takes 98% of the championships. With twenty-two, 100%. Making the year seventeen races longer changes the champion in two seasons out of a hundred.
2010 says the same with a nice wrinkle: over five races there are six different champions across 150 runs — Webber, Hamilton, Button, Massa and even the control driver get one — but Vettel already wins 91%. The variety lives in the 9% underneath; at the top there is no argument even with a minimal calendar.
Put another way: when one car is clearly better, the championship is settled before it starts and the races only make it official. Exactly how much that car weighs is measured separately in how much the car matters in Formula 1.
In a close year, twenty-two are not enough
Now 2007, the exact reverse. Over five races the favourite wins 49% — less than half: it is literally a coin flip between Hamilton and the rest of the world. Over ten it climbs to 66%. And from there it stops climbing: 67% over sixteen, 67% over twenty-two.
Twelve more races, exactly zero percentage points gained. A full Formula 1 season, with all its logistics and its 22 weekends, and in a third of the universes somebody else takes it. Not because the system is broken: because in 2007 the top three were worth practically the same, and when two things are worth the same, repeating the measurement does not separate them.
The champion’s margin says it plainly: 7.9% in 2007 against 33% in 2024. In 2007 the title is won by the equivalent of two good races; in 2024, by eight.
The impossible-calendar experiment
If twenty-two races are not enough, would forty be? We ran it: 150 seasons of 2007 and 150 of 2021 with a 40-race calendar, nearly double the longest season in real history.
- 2007 goes from 67% to 81%. It climbs — at the price of doubling the championship.
- 2021 does not move: 88% over twenty-two races, 85% over forty. Within the sampling error, nothing changes.
That is the answer to the question in the title, and it is awkward for both sides of the argument. A longer calendar does not make a fairer championship: it makes a longer championship. What decides whether the best driver wins is not in the calendar, it is in the technical regulations and in the gap between one car and the next.
Different champions, the other side of the same coin
The same measurement told backwards: how many different drivers took the title across the 150 runs of each cell.
| Season | 5 races | 10 | 16 | 22 |
|---|---|---|---|---|
| 2007 | 4 | 4 | 3 | 3 |
| 2010 | 6 | 3 | 2 | 1 |
| 2021 | 2 | 2 | 2 | 2 |
| 2024 | 2 | 2 | 1 | 1 |
A short calendar does not change who the favourite is: it changes how many people get a shot. In 2010 over five races there are six possible names; over twenty-two there is one. That is a fairly exact way of saying what a long season does — it does not pick the champion better, it eliminates the outsiders.
And the promised detail: in the five-race 2007 championships, our control driver — an average driver, nothing special about him, sitting in the second Ferrari — took 4% of the titles. Over ten races it drops to 3%. Over sixteen he disappears. The lottery of short championships is not a figure of speech: it has a number, and it is small but not zero.
And in real Formula 1, did that happen?
This is where we should be honest about what this is: a calibrated model, not a crystal ball. But the contrast with what actually happened is striking, and we never asked the engine to reproduce it.
- 2007 was decided by one point between three drivers, with the title fight alive until the last lap of the last race. Our model says that year had no clear owner even with double the races.
- 2010 reached the final round with four mathematical contenders and Vettel won it — exactly the favourite that comes out here, winning 91% with just five races.
- 2021 was settled on the last lap of the last grand prix. In the simulation it is the only one of the four seasons that keeps gaining certainty with every race you add: it is a duel, and duels resolve slowly.
- 2024 was closed out with four races to spare. In the simulation, five races are enough to settle it.
What the numbers say
When a championship is settled early and the call for more races comes back, the numbers give it little support: in a lopsided year extra races change nothing, and in a close one the calendar has to double to move fourteen percentage points. What makes a championship exciting is not its length, it is the distance between the first car and the second.
Individual scenarios can be built in What if? — championship length included — and the pre-computed verdicts cover the most requested ones. The other side of the coin, how much of the result is the driver and how much the machinery, is in talent or car.