2022 Mathematical Driver Rankings

Greetings! As I did last year, I am presenting my mathematical model’s driver rankings for 2022. For those unfamiliar with the model, it compares drivers against their teammates across their entire career to output values for how good they are. It then compares an expected result to the actual result obtained each year to rank drivers. You can read more about it here.

The rankings and ratings are based on the exact same methodology as last year. However, I have also calculated how non-driver related DNFs affect the scores. As this was started after the season commenced, I have added it as an adjustment at the end of the article. However, I intend to use this system going forward.

  1. Performance within each team
  2. D Tier
  3. C Tier
  4. B Tier
  5. A Tier
  6. S Tier
  7. Rankings Including DNFs

Performance within each team

First, let’s start with a quick look at how each driver performed versus their main teammate, in terms of points scored:

*de Vries and Hülkenberg are not included here.

There are some fairly dominant teammate match ups for Bottas and Norris, whilst Mercedes and Alpine had two of the closest pairings on the grid. You can also see that more competitive cars are likely to be closer in terms of % points (but not actual points), due to the large number of points on offer. The difference in the likely percentages of teammates of teammates is something that the model accounts for.

We can also see a final standing of how the models’ preseason predictions have held up:

As usual it’s a bit of a mixed bag, with the main discrepencies being at Williams (which was predicted at the time) and McLaren (which wasn’t). The model is tending towards the right answer (compared to 2021) for Red Bull, Ferrari, AlphaTauri and Aston Martin, whilst for Alpine and McLaren the results tended against the prediction. (Alonso and Ricciardo were expected to do better in 2022 vs their teammates, but instead they did worse.)

The graphs also quickly demonstrate which drivers have performed better or worse than the model’s predictions. Due to Alfa Romeo featuring a rookies and Haas driver Schumacher not being linked to the rest of the grid, predictions of their ratings were not made. For what it’s worth, I’ll be posting the 2023 Driver predictions at the start of January.

Onto the driver ratings! Remember that a score of 100% is considered to be an average score for Hamilton across this career, whilst other scores are reflective of how many points the model would expect this performance to earn relative to Hamilton in a competitive car (e.g. a score of 70% would mean 7 points for every 10 Hamilton scored in the same car).

Each driver has been given a ranking for the grid (1-20), a % score, and a number in brackets when applicable that shows if their rankings has improved or gone down since last year. Finally, as the ratings can be quite erratic in the first few races, the graphs show the ratings how changes from the 6th Grand Prix onwards.

D Tier

20) Zhou Guanyu, 49% (N/A)

When Zhou was announced as Alfa Romeo’s second driver, much of the focus was on his nationality and the financial benefits he brought to the team. Given this, he has probably exceeded expectations, and earnt a drive for 2023 to fully prove his worth. So why does the model rank him so poorly?

Well, the output is based on results and points. Going into the season, I suggested that scoring roughly half of Bottas’ points was a realistic aim for the rookie. Given that, an output of just 6 points is considered poor, regardless of potential mitigating circumstances. The Alfa Romeo points totals are the most imbalanced on the grid, after all.

In his defense, almost all of Bottas’ points were delivered at the start of the year, when Zhou’s lack of experience really showed. Zhou developed as the season progressed, which can be shown by a steady rise in his rating, even if the points scoreboard doesn’t reflect that. Finally, rookies have also often struggled to deliver in their first year (Mazepin, Schumacher and Tsunoda are all near the bottom of ranked drivers for 2021, for example). There have been signs this year that he can deliver, but he’ll need to score more points relative to his teammate next year to show he can turn potential into reality.

19) Mick Schumacher, 55% (0)

Schumacher’s debut year against Mazepin was encouraging, and the expectation was that a year’s experience would help him become a more consistent driver. The reality was quite different, with his tendency to crash the car still not fully ironed out, and his perceived speed from last year far less impressive when pitted against a known benchmark like Magnussen. Two consecutive points finishes in the middle of the season (seen in the graph below as a jump in his rating) seemed like a breakthrough, but it never fully carried through to the rest of the year.

Like Zhou, Schumacher was comfortably outscored by a teammate that is not considered top tier in Formula One. It’s now looking like Schumacher will be out of Formula One, and it’s an open question whether it’s a good move by Haas or not. As a rule of thumb, a score of 50-60% is on the bubble of when driver’s typically lose their drives (Giovinazzi scored 56% last year, for example), although rookies are often given a bit more leeway. Given this, it was reasonable for Haas to be looking elsewhere.

18) Kevin Magnussen, 62% (N/A)

A whirlwind season for Magnussen, whose low ranking will ire some. High points included returning to the sport at all, producing a fantastic 5th place in the season opener and an even more impressive pole lap in Brazil. The Dane often produced feel good stories throughout the year, and reestablished himself on the grid.

The model, however, is not particularly complimentary of Magnussen’s overall F1 career. This, combined with the fact that Schumacher has no prior match ups with a driver connected to the rest of the grid, puts him lower down on this list than many would expect. 2023 will serve as a great lesson in whether he’s better than the model currently gives him credit for.

Magnussen's ranking was static all year, due to his rookie teammate. Schumacher's jumped mid-season with points finishes, whereas Zhou's inched upwards all year.

4 thoughts on “2022 Mathematical Driver Rankings

  1. Acescheil

    That really makes me curious to see how taking into account DNF/DNS will change McLaren 2015, Raikkonen 2005 and if you ever will go as far back as redoing some seasons in the ’70 and earlier.
    And one question: if you are gonna do some of the rankings of the previous season accounting for reliability, will you need to do the ranking based on that season alone? Or will you also take into account for the ranking the performance of the driver in the other seasons, even if the other seasons don’t take reliability into account?

    Like

    Reply
    1. formula1stat Post author

      I’m also intrigued about Räikkönen in the McLaren and some of the older years! You’ve identified a weakness of the DNF approach for 2022, that it takes every season into account but only DNFs in 2022.

      The idea is to go back through every season, although it will take some time as a lot of the input will be manual. Luckily we’re in the off-season. The aim is to have it all up and running by the start of the 2023 season, or failing that, to have it go back far enough so that modern seasons will be minimally affected by any further changes.

      Like

      Reply
      1. Acescheil

        I’ve got 2 questions:
        1. How are you dealing with DNFs when the driver completed at least 90% of the race and therefore is still classified?
        2. If you’re answer to the first question was ‘I count them as proper DNFs’ then how are you going to deal with the situations in which a driver retired, but completed enough laps to be classified AND classified in a point scoring position?
        I was thinking about a solution to the second would be to count the situations as DNFs and also taking away the points they scored in that race to the total at the end of the year, but I don’t know if you’re can do that without rewriting a huge chunk of the algorithm

        Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s