Why it is so hard to compare racing drivers across series

Motorsport has no equivalent of a batting average. A driver who wins ten club races and a driver who finishes eighth in Formula 1 have records that cannot be read side by side, and there is no governing body that publishes a number saying which of them is faster.

This index holds 49,966 races across 138 series. Building one rating that spans them meant solving four problems first. None of them has a clean answer.

Problem one: a win means completely different things

The most obvious measure is the worst one.

Take the median rating of every driver who started a race in each series since 2024. That gives a rough sense of who you are racing against.

The gap between the top and bottom of that table is close to 6,000 points. A win at one end of it and a win at the other are the same word for two different achievements.

Any rating that counts wins equally is broken before it starts. What matters is who you beat.

Problem two: most drivers never race each other

Ranking systems work by connection. You can rank two tennis players who have never met if they have both played enough people in common.

Motorsport is badly connected. A club racer in Alabama and a Formula 1 driver share no opponents, no season, and usually no era. Whole series sit almost entirely disconnected from the rest of the sport, and the ones that connect do so through a handful of drivers who moved between them.

Those few crossover drivers carry an enormous amount of weight. When a Formula 2 driver moves to IndyCar, or a club racer steps up to a national GT series, their results are what pins one scale to the other. Get those drivers wrong and two entire championships end up misplaced relative to each other.

This is why a rating cannot be computed series by series and then stitched together. It has to be computed across everything at once.

Problem three: classes inside one race

A GT field can have four classes on track together. The car that finishes tenth overall might have won its class by a lap.

Scoring that by overall position is wrong, because a GT4 car is not racing the prototype in front of it. Scoring it by class position alone is also wrong, because "first in class" against three cars is not the same as first in class against twenty.

This one bites in a specific way. When we first rated IMSA WeatherTech, GTD Pro and GTP were being treated as one field, which made GTD Pro look far stronger than it was. Splitting them changed the ratings of every driver in both classes.

Getting class handling wrong does not produce small errors. It produces confidently wrong answers about who is quick.

Problem four: shared cars

Endurance racing puts three or four drivers in one car and gives them one finishing position.

If a car wins Sebring, which of its drivers was fast? The result cannot tell you. It might be that all three were quick, or that one was quick and two were adequate, or that the car was quick and none of them mattered much.

Our convention is that every driver of a car takes the car's result, which is the same convention the championships use. It is not satisfying. A driver who was carried by a stronger co-driver gets credit they did not earn, and it takes several seasons of other results to correct.

Where per-driver stint data exists it is much better, and we use it where we have it. For most endurance racing it does not exist.

What we ended up doing

The Racer Rating is an Elo-style number. Everyone starts from a level appropriate to where they are racing, and every result moves them based on who they finished ahead of and behind, weighted by how strong those drivers are.

Beat someone rated 6,000 and you gain a lot. Beat someone rated 1,200 and you gain almost nothing. Win ten club races against the same twelve people and your rating stops moving, because you have already proved you are faster than them.

That last property is the one that makes the whole thing work. It means a rating cannot be farmed by racing weak fields, which is what makes a club champion and a professional comparable at all.

What it still gets wrong

A driver who moves between very different disciplines gets penalised for the adjustment period. The rating reads that as a loss of pace. It is actually a change of job, and we have not fixed it.

Coverage is uneven. Some series are indexed back twenty years and some only from 2022, so a driver's early career can be missing entirely.

And a rating is a measure of results, not of talent. A quick driver in poor equipment rates lower than a slower driver in good equipment, and no amount of maths gets around that.

---

Read the full method on the Racer Rating page, or start with the driver index.