2026 is offering a lot of challenges for anyone trying to make sense of driver performances. The new regulations seemingly don’t reward pushing to the limit the same way as before, and quite a few drivers are struggling to adapt. Furthermore, a sizeable portion of the grid is yet unconnected, leaving most teammate duels up to interpretation. Due to these issues, this will not be a straightforward list, but more of a discussion of possible outcomes. The mathematical model seeks to eliminate the factor car by only measuring drivers against their teammates.
How Does It Work?
My model was inspired by the amazing f1metrics model and therefore shares a significant portion of its methodology.
Step 1 was to introduce a unified points system. P1 gets 10p, P2 gets 10p*0.72=7.2p, P3 gets 7.2p*0.72=5.18p …. etc. all the way to P20. Driver related DNFs are scored with 0p.
Then the 2 teammates get their points assigned for every race. Races where one driver has mechanical issues are excluded (or corrected if possible). Driver errors and other incidents are also excluded (or corrected if possible). Now the points are tallied up and converted on an s-curve. The difference between the two values of the drivers is the performance difference between them.
In combination with functions for age and experience (I use those of the f1metrics model), doing this with all drivers over a lot of seasons, results in a pretty sturdy web of relative driver performances. While building this web, I am not looking for a compromise of all seasons but instead a most perfect fit.
Season Ratings
Now that I have performance ratings (later referred to as ceiling/pace/potential) for the drivers, I can get a value for the car/team in a season since
(points)=(driver performance)+(car performance)
and therefore
(car performance)=(points)-(driver performance)
The driver errors are then added back in, giving a new points total, and through
(points)-(car performance)=(driver performance)
I get the actual performance of the drivers in that season. Those performance ratings are then translated back over the s-curve to give a “points per race rating” which goes from 0-10. That is then multiplied by 10 to get a neat driver score of 0-100.
As you can see, the major difference to the f1metrics model is the handling of driver errors. In my opinion this results in more accurate ratings as the amount of errors a driver makes can vary widely from season to season. Secondly, it has the advantage of untieing teammates in a season. A drivers rating won’t automatically get boosted just because his teammate is having a shocker.
At its core the model is assessing a drivers ability to score points with
(total points)=(possible points)-(lost points)
This can be abstracted onto the drivers. Simplified the drivers’ season ratings can then be understood as
(driver rating)=(driver pace)-(driver mistakes)
To help give some perspective for the ratings I made these tiers:
- S – above 89
- A – above 82
- B – above 73
- C – above 62
- D – above 50
- F – below 50
From going through all the data and seasons, I would say that generally a good season has a score above 73. Keep in mind, though, that there are no real cut lines. Another bit of context is that the median performance usually sits somewhere in the 60s. Over the last couple of years, it has crossed over into the 70s, which is the first time this happened since the 70s.
The “Ratables”
These are the double anchored teammate pairings, where both drivers have already enough data to support a ceiling. Ferrari’s, McLaren’s, Williams’, Cadillac’s, and Aston Martin’s line-ups are relatively unaffected by the uncertainty brought through the recent influx of rookies.
Aston Martin
Unfortunately, Aston Martin had too many problems with their car to generate a meaningful sample of clean races, meaning they are infact not ratable after all. Fernando Alonso’s current potential is 79.4, according to the model. For anyone convinced that he has been flawless so far inspite of the difficult situation, that would be his rating. Lance Stroll’s current potential (and ceiling) is 74.1.
Fernando Alonso / (79.4)
Lance Stroll / (74.1)
Cadillac
Cadillac also have had massive issues with reliability; as much is expected from a brand new team. What wasn’t expected, however, was the delta over the few clean races they had. Perez holds a healthy lead over Bottas, when their previous work suggested a balanced pairing. Of course, it is not known whether or not their races have been representative or mostly glorified test sessions. At least as likely, though, Bottas is simply underperforming; just like many others are. Perez on the other hand lost a good chunk of his rating in Monaco due to his silly repeated start infringements.
Sergio Perez 55.2 (68.5)
Valtteri Bottas 52.7 (68.5)
Williams
At the moment, it looks like Carlos Sainz has the measure of Alex Albon on pace. The qualifying record is clear (9-1!), and he is narrowly ahead on points going into the summer break (6-5). Sainz having the upper hand in qualifying isn’t new, as he has become more and more dialed towards Saturdays over the last 5 years. However, even on race day Albon usually trails his teammate in 2026, albeit much closer than over one lap. The crucial exception to all of this was the Monaco Grand Prix, where Alex outqualified Carlos and lead him until an unnecessary collision with Hulkenberg meant that the Spaniard lost out on the biggest points scoring opportunity for Williams in this disappointing season. This resulted in a big hit to his rating. Without this single race, the ratings would be 80 to 68 in Sainz’ favour.
Alex Albon 76.4 (83.2)
Carlos Sainz 74.9 (82.8)
Ferrari
Going into the season, the model anticipated basically a repeat of last year. This has not been the case. Although the delta is on Leclerc’s side, it sits far beneath expectation. While it is easy to get swept up in a good comeback story, attributing this new-found competitiveness to Hamilton somehow suddenly clawing back what a decade of aging took from him would be unreasonable. It is also doubtful for this deficit to stem from Leclerc’s brake change because removing Canada through Barcelona does not move the needle nearly enough. More likely, Leclerc, like Russell and probably a few more, hasn’t quite gotten his head around the new cars. Rarely has Charles struggled for pace like he has at times this year. It will be interesting to see how long it takes him to fully get on with it. Hamilton, in turn, seems to have clicked with the 2026 regulations much more quickly. Overall, the pair is incredibly closely matched on performance, owing the sizeable points gap to Leclerc’s 2 mechanical failures, while Hamilton remains the only driver to finish every race. This would be a fantastic set-up if they had more than just the best chassis under them.
Charles Leclerc 80.2 (92.1)
Lewis Hamilton 80.1 (82.5)
McLaren
Contrary to current discourse, Norris’ delta to Piastri is not larger than expected. Mechanical issues might skew the picture slightly, bringing down the sample size, however, both drivers have taken home big results over the other, and both have missed out on favourable outcomes due to reliability. Norris has been more unlucky overall, though, with especially the Belgian Grand Prix seeming like a major lost opportunity. Part of the explanation might simply be that the model didn’t expect them to be equal on pace to begin with. To the model, Norris is and always has been the better driver with a slightly higher ceiling even with both in their prime. If it is more than that and indeed just a matter of sample size, Piastri’s underperformance will show up sooner or later, provided he is actually underperforming. For now, this leads to both McLaren drivers settling at the very top of the limited number of ratable drivers; though with Lando clearly ahead.
Lando Norris 88.7 (89.3)
Oscar Piastri 84.3 (87.0)
The “Speculatives”
These are the problematic cases. The combination of the aforementioned different factors leaves a lot of room for interpretation through different permutations; if there is even anything to go off of that is. They range from “actually pretty likely correct” all the way to “nothing there”.
Red Bulls
The Red Bull block presents a complex situation. All of Hadjar, Lawson, and Lindblad are unknowns as far as the model is concerned due to none of them having at least 2 matching data points. This leads to Verstappen and Hadjar being just tethered to what’s expected of Max by the model for the current ratings, meaning there is no counterweight to catch it, should Verstappen underperform. With Racing Bull’s line-up on the other hand, there is nothing at all; the model can not rate them at this point.
While Lindblad is completely without data, Lawson already has a delta against Tsunoda and Hadjar. The Tsunoda comparison has only a handful of races, making the range quite big, but if it is assumed representative, Lawson’s and Lindblad’s ratings would be in the ranges of 71-61 and 64-53, respectively. I already used the possible lower Tsunoda-Lawson delta for the 2025 rating estimates of Hadjar and Lawson, and as it stands, the 2026 delta between Verstappen and Hadjar would fit right in with that. Should this hold true until the end of the season, Hadjar and Lawson would get a ceiling. It would also imply that Verstappen is likely not underperforming. Assuming the current delta between Lawson and Lindblad is also representative, their ceilings would look like this:
- Hadjar (82.1)
- Lindblad (77.7)
- Lawson (73.6)
Unsurprisingly, Hadjar clearly has a higher ceiling than Lawson, as Red Bull would agree, and while Lawson is currently beating his teammate relatively comprehensively, Arvid’s age and level of experience easily allow him to out-develop Liam. Again, that is if their current delta proves to be representative. This would also mean that Red Bull and Racing Bulls have very similar cars in terms of overall performance.
Max Verstappen 91.7 (92.9)
Isack Hadjar 68.3 (*70.1)
Liam Lawson 61.2 (*66.3)
Arvid Lindblad 53.7 (*53.7)
Alpine
Gasly and Colapinto are basically in the same position as Verstappen and Hadjar, where everything is based on the main driver due to his teammate not being properly connected. However, unlike with Red Bull, the delta between the Alpine drivers is not yet clearly converging towards a specific area. Their 2025 delta would put the expectation for 2026 right in the middle of the current range, though, which is a good sign. If it does end up matching, Colapinto’s ceiling would predict him to become a Gasly-level driver himself, and to have a potential of 71 in 2026. As this is not yet the case, their ratings should be viewed with extra caution. Assuming, Gasly is not fundamentally underperforming, he has been doing a fantastic job so far, and, according to the model, needs to be mentioned in the same breath as the Ferrari drivers for the first half of the season.
Pierre Gasly 80.2 (80.2)
Franco Colapinto 60.7 (*70.6)
Audi
Once again, this presents a similar situation to the previous entry with just a bit more uncertainty. The delta between Hulkenberg and Bortoleto is actually bigger than what their 2025 season would predict. This could be taken as evidence that Hulkenberg is at least not underperforming, but it might just as well be that he severely underperformed last year. Additionally, mechanical gremlins have taken their toll on Audi’s debut season (as you would expect), meaning sample size could play a role here too. In any case, it’s simply unknowable at the moment, giving no choice but to trust Hulkenberg to be performing as expected until more data is available.
Nico Hulkenberg 61.0 (65.7)
Gabriel Bortoleto 56.4 ( / )
Mercedes
At Mercedes, a much more intriguing situation is unfolding, because the known quantity is clearly not delivering based both on the numbers and George Russell’s own words. If he were, his and Kimi’s ratings would be 81.0 and 92.3, respectively. Not only would that make Antonelli better than Verstappen so far, it would also mean that this 19-year-old sophomore must have a ceiling beyond reality, a ceiling which would make prime Hamilton look like Bottas next to him.
Obviously, Antonelli doesn’t have a ceiling yet, but based on his rookie season his rating would be expected in the range of 72-78; from which it follows, that Russell sits at 48-56. Even with all caveats, these numbers are likely a truer representation of their performance than the current actual ratings. Kimi is most likely not in the running for the best driver of the season in 2026 (and how could you expect him to be as a 19-year-old with a single previous season of F1-experience). George on the other hand is genuinely in trouble to be one of the worst performing drivers of 2026. He seems absolutely lost compared to his teammate. Vettel was in a similar spot in 2014. While he was clearly unlucky, he also just wasn’t able to get his best pace out of the car unlike Ricciardo. Vettel ended up having a mediocre season, which seems like the best Russell can be hoping for right now. Terrible timing, though.
George Russell 81.0 (85.5)
Kimi Antonelli 92.3 ( / )
Haas
Look to the 2025 driver ratings to see a breakdown of what might be going on around Ocon (and therefore Bearman).
2026 has brought another layer on top of what was already a confounding situation, because Ollie Bearman is beating Esteban Ocon by much more than even last year’s delta, which already suggested an Ocon underperformance, would predict. Is Ocon underperforming his underperformance? Is Ocon generally overrated by the model and now underperforming? Was Bearman underperforming in 2025? Or is Bearman God’s gift to motorsports? This pairing leaves enough room for interpretation to bring a literatur student to tears. Yet, it’s pointless. There are no ranges or likely scenarios to put this one in any sort of reasonable light.
Esteban Ocon 73.6 (81.4)
Oliver Bearman 87.6 ( / )
F1 Mathematical Model 
Leave a comment