Beleriand's Statistical Corner

Forum rules
Please read the forum rules
Beleriand_K
Posts: 25
Joined: Wed May 20, 2020 2:15 pm

Beleriand's Statistical Corner

Post by Beleriand_K »

The official Formula 1 drivers standing is to a very high degree decided by the machinery, not the driver. According to Nico Rosberg earlier this year, the car accounts for approximately 80% of the result and the driver for only 20%.

Even though the best drivers are usually hired by the best teams, the cars massive influence on the result can still create a significant error in the official standing. One obvious example was Fernando Alonso finishing #11 in the final 2018-standing behind Carlos Sainz, Kevin Magnussen and Sergio Perez. I think most fans will agree this standing painted a very flawed picture of Alonsos strength.

It's impossible to split the results in two parts, so the drivers part of it stands out clear and undisputable. But that doesn't mean that it's impossible to have an opinion about it. And it doesn't mean, that this opinion can't be based on the available facts. That's where ratings come into the equation.

Even though ratings of drivers strength isn't the undisputable truth (neither is the official standing), they often provide alternative ways of estimating the drivers "real" position in the Formula 1 hierarchy of strength. And new ways of looking at thing is very often interesting.

Enough talk. Let's get ready to rumble...

I like numbers, statistics and Formula 1, so I've made an alternative standing based on available information about races, drivers and teams, and when this is exposed to a few statistic tools the result is a Drivers Strength Standing adjusted for the cars estimated contribution to the results.

Please notice the word "estimated", as the following rating doesn't claim to be a scientific 100% correct rating. Such a rating doesn't exist. But I believe it does come a long way towards an unbiased rating of the drivers current strength.
Top 20%Max Verstappen64.9
Charles Leclerc62.0
Lewis Hamilton60.7
Sebastian Vettel58.2
Inside Top 10Valtteri Bottas54.8
Kimi Raikkönen54.5
Sergio Perez52.6
Daniel Ricciardo52.1
Carlos Sainz Jr.50.1
Daniil Kvyat47.3
Outside Top 10Lando Norris47.2
Romain Grosjean46.4
Alex Albon46.1
Nico Hülkenberg42.2
Kevin Magnussen42.1
George Russell41.5
Bottom 20%Pierre Gasly39.8
Antonio Giovinazzi38.4
Lance Stroll37.1
Robert Kubica22.8
A few comments to the rating:
1) I believe that Lando Norris will pass Daniil Kvyat during the upcoming season, and that the reason for Kvyat currently being slightly ahead is experience.

2) I also believe we will see a driver like George Russell move upwards in the strength hierachy as he gains experience. His position in the rating might be less interesting than the distinct difference between him and teammate Robert Kubica, who unfortunately didn't succeed with his attempted comeback.

3) It surprised me a little that both Verstappen and Leclerc has passed Hamilton, but it's a close fight at the top.

4) Ferrari has just let the fourth strongest driver go and replaced him with the ninth strongest. It doesn't seem to make sense, but remembering Toto Wolffs words about the importance of a Wingman it might make sense anyway.

User avatar
Invade
Posts: 3053
Joined: Mon Oct 26, 2015 9:33 pm

Re: Beleriand's Statistical Corner

Post by Invade »

I'm suspicious of some of the findings which appear counterintuitive to common sense and judgment that can be made through simple observation, but I am very intrigued by your project and would love to hear more about it.

Beleriand_K
Posts: 25
Joined: Wed May 20, 2020 2:15 pm

Re: Beleriand's Statistical Corner

Post by Beleriand_K »

Invade wrote:I'm suspicious of some of the findings which appear counterintuitive to common sense and judgment that can be made through simple observation, but I am very intrigued by your project and would love to hear more about it.
I understand the way you feel about it, and I felt pretty much the same way, when the results started popping up. For instance, I would put Ricciardo above both Bottas, Raikkönen and Perez, but when we look at the actual results during the 2019-season, there is simply no ground for doing that.

It's important to remember that this is the current strength standing. Not an overall rating of the drivers qualities during a lifetime in Formula 1. I also have a standing showing the average ranking for the past three seasons (Hamilton #1 followed by Verstappen and Vettel), but the ranking above is based on the last 20 races. So if a strong driver like Ricciardo for whatever reason has a mediocre season (as he did in 2019), then he moves down the short term strength list. Which is just the way it should be.

mikeyg123
Posts: 17831
Joined: Sat Jan 28, 2012 4:13 pm

Re: Beleriand's Statistical Corner

Post by mikeyg123 »

Beleriand_K wrote:
Invade wrote:I'm suspicious of some of the findings which appear counterintuitive to common sense and judgment that can be made through simple observation, but I am very intrigued by your project and would love to hear more about it.
I understand the way you feel about it, and I felt pretty much the same way, when the results started popping up. For instance, I would put Ricciardo above both Bottas, Raikkönen and Perez, but when we look at the actual results during the 2019-season, there is simply no ground for doing that.

It's important to remember that this is the current strength standing. Not an overall rating of the drivers qualities during a lifetime in Formula 1. I also have a standing showing the average ranking for the past three seasons (Hamilton #1 followed by Verstappen and Vettel), but the ranking above is based on the last 20 races. So if a strong driver like Ricciardo for whatever reason has a mediocre season (as he did in 2019), then he moves down the short term strength list. Which is just the way it should be.
With the top 5 drivers all coming from the top 3 teams are you confident you have the balance between car and driver correct? It looks to me, with that and George Russell's low placing that you are not correcting for the car enough?

When assessing the value of the car are giving it a score per race or over the whole season? If it is the later then that would account for Raikkonen's high score. He scored a lot of points early on in the season when the Alfa was very competitive but hardly any later in the year when it had probably slipped to 8th best.

How are you establishing what score to give the scores? If it is the speed they are being driven it or results they are achieving is that not somewhat manipulated by drivers as well? A pair of really good drivers will make a car look better than a pair of average ones for example.

User avatar
Alienturnedhuman
Posts: 3899
Joined: Thu Feb 13, 2003 9:39 pm

Re: Beleriand's Statistical Corner

Post by Alienturnedhuman »

I think a key thing to ask here is what exactly are you doing? Are you literally taking Nico Rosberg's comment as "the car is responsible for 80% and the driver 20%" at face value and doing a split based on that ratio. If so then this is going to be completely meaningless because wordings like that are never meant to be literal, it's just a way of saying "it's heavily weighted to towards the car, but the driver can make an impact" - and even if it could be quantified like that, a perfect 1:4 ratio of driver to car seems awfully tidy and convenient to have any accuracy...

I think you need to explain how you've arrived at these figures in a little more detail in order for us to assess them.

Beleriand_K
Posts: 25
Joined: Wed May 20, 2020 2:15 pm

Re: Beleriand's Statistical Corner

Post by Beleriand_K »

mikeyg123 wrote: With the top 5 drivers all coming from the top 3 teams are you confident you have the balance between car and driver correct? It looks to me, with that and George Russell's low placing that you are not correcting for the car enough?

When assessing the value of the car are giving it a score per race or over the whole season? If it is the later then that would account for Raikkonen's high score. He scored a lot of points early on in the season when the Alfa was very competitive but hardly any later in the year when it had probably slipped to 8th best.

How are you establishing what score to give the scores? If it is the speed they are being driven it or results they are achieving is that not somewhat manipulated by drivers as well? A pair of really good drivers will make a car look better than a pair of average ones for example.
I would love to say that "Yes, I'm sure, I have the balance between car and driver correct". But that would not be true.

In the real world the balance between the car and driver is an unknown parameter, that has to be estimated. I have used the 80/20 suggested by Nico Rosberg, but I have read another analysis claiming the balance to be more like 85/15 today falling from 70/30 some decades ago.

The scores are based on a drivers average laptime during a race compared with the average laptime for all the drivers completing the race. This difference is adjusted for the cars average advantage/disadvantage compared with the other teams during the season. During the last 20 races Mercedes preformed on average 1,32% better than a "neutral" team, while Williams on average performed 1,63% worse. It is well known that some cars perform better on some tracks than other, but by including 20 races this should be equalled out.

It is true that two really good drivers can make a car look better than two average drivers, but I believe the law of supply and demand will equal that out, as the best drivers will be snapped up by the best teams, while the average drivers will be stuck with the less attractive teams.

Regarding George Russel I mentioned him in a comment, and he would be placed higher with more weight put on the cars importance. But on the other hand we have to remember, that he is an unexperienced driver, and experience is not an unimportant part of a drivers strength. I think the big gap between him and Robert Kubica says a lot about George Russel, and I certainly expect him to advance in the ranking, when we finally get the season started.

mikeyg123
Posts: 17831
Joined: Sat Jan 28, 2012 4:13 pm

Re: Beleriand's Statistical Corner

Post by mikeyg123 »

Beleriand_K wrote:
mikeyg123 wrote: With the top 5 drivers all coming from the top 3 teams are you confident you have the balance between car and driver correct? It looks to me, with that and George Russell's low placing that you are not correcting for the car enough?

When assessing the value of the car are giving it a score per race or over the whole season? If it is the later then that would account for Raikkonen's high score. He scored a lot of points early on in the season when the Alfa was very competitive but hardly any later in the year when it had probably slipped to 8th best.

How are you establishing what score to give the scores? If it is the speed they are being driven it or results they are achieving is that not somewhat manipulated by drivers as well? A pair of really good drivers will make a car look better than a pair of average ones for example.
I would love to say that "Yes, I'm sure, I have the balance between car and driver correct". But that would not be true.

In the real world the balance between the car and driver is an unknown parameter, that has to be estimated. I have used the 80/20 suggested by Nico Rosberg, but I have read another analysis claiming the balance to be more like 85/15 today falling from 70/30 some decades ago.

The scores are based on a drivers average laptime during a race compared with the average laptime for all the drivers completing the race. This difference is adjusted for the cars average advantage/disadvantage compared with the other teams during the season. During the last 20 races Mercedes preformed on average 1,32% better than a "neutral" team, while Williams on average performed 1,63% worse. It is well known that some cars perform better on some tracks than other, but by including 20 races this should be equalled out.

It is true that two really good drivers can make a car look better than two average drivers, but I believe the law of supply and demand will equal that out, as the best drivers will be snapped up by the best teams, while the average drivers will be stuck with the less attractive teams.

Regarding George Russel I mentioned him in a comment, and he would be placed higher with more weight put on the cars importance. But on the other hand we have to remember, that he is an unexperienced driver, and experience is not an unimportant part of a drivers strength. I think the big gap between him and Robert Kubica says a lot about George Russel, and I certainly expect him to advance in the ranking, when we finally get the season started.
I appreciate the work you are doing and it is interesting. I just don't think these things can mathematically calculated. The driver is obviously influenced by the car but the car is also influenced by the driver. Mercedes had better drivers than Haas but the stats you are using to get the cars baseline assumes both pairings are equal if you see what I mean.

Herb
Posts: 2352
Joined: Wed Dec 04, 2013 4:42 pm

Re: Beleriand's Statistical Corner

Post by Herb »

Beleriand_K wrote:
mikeyg123 wrote: With the top 5 drivers all coming from the top 3 teams are you confident you have the balance between car and driver correct? It looks to me, with that and George Russell's low placing that you are not correcting for the car enough?

When assessing the value of the car are giving it a score per race or over the whole season? If it is the later then that would account for Raikkonen's high score. He scored a lot of points early on in the season when the Alfa was very competitive but hardly any later in the year when it had probably slipped to 8th best.

How are you establishing what score to give the scores? If it is the speed they are being driven it or results they are achieving is that not somewhat manipulated by drivers as well? A pair of really good drivers will make a car look better than a pair of average ones for example.
I would love to say that "Yes, I'm sure, I have the balance between car and driver correct". But that would not be true.

In the real world the balance between the car and driver is an unknown parameter, that has to be estimated. I have used the 80/20 suggested by Nico Rosberg, but I have read another analysis claiming the balance to be more like 85/15 today falling from 70/30 some decades ago.

The scores are based on a drivers average laptime during a race compared with the average laptime for all the drivers completing the race. This difference is adjusted for the cars average advantage/disadvantage compared with the other teams during the season. During the last 20 races Mercedes preformed on average 1,32% better than a "neutral" team, while Williams on average performed 1,63% worse. It is well known that some cars perform better on some tracks than other, but by including 20 races this should be equalled out.

It is true that two really good drivers can make a car look better than two average drivers, but I believe the law of supply and demand will equal that out, as the best drivers will be snapped up by the best teams, while the average drivers will be stuck with the less attractive teams.

Regarding George Russel I mentioned him in a comment, and he would be placed higher with more weight put on the cars importance. But on the other hand we have to remember, that he is an unexperienced driver, and experience is not an unimportant part of a drivers strength. I think the big gap between him and Robert Kubica says a lot about George Russel, and I certainly expect him to advance in the ranking, when we finally get the season started.
I'd love to see the workings of this.

But regarding the bit I have highlighted.

Doesn't this disadvantage the faster drivers/teams?

Once a driver is in the lead, there is no point in pushing the car more than necessary. They also need to make the car parts last as long as possible, so it is well known that drivers will sometimes take their foot off the gas - leading to a slower average lap time. This may not be possible for the cars in the midfield, that are often in a dogfight for large parts of the race.

Beleriand_K
Posts: 25
Joined: Wed May 20, 2020 2:15 pm

Re: Beleriand's Statistical Corner

Post by Beleriand_K »

Herb wrote:Once a driver is in the lead, there is no point in pushing the car more than necessary. They also need to make the car parts last as long as possible, so it is well known that drivers will sometimes take their foot off the gas - leading to a slower average lap time. This may not be possible for the cars in the midfield, that are often in a dogfight for large parts of the race.
I think the problem with having to take the foot of the gas for technical reasons hit everybody once in a while. And the midfield cars are not necessarily under more pressure than the top cars. Kevin Magnussen said once after a race, that it had been very boring, because he was far behind the top 3 teams, but strong enough to keep a safe distance to the rest of the field. So he had just finished lap after lap pretty much on his own. That was once upon a time, when Haas was a fast car...

User avatar
Alienturnedhuman
Posts: 3899
Joined: Thu Feb 13, 2003 9:39 pm

Re: Beleriand's Statistical Corner

Post by Alienturnedhuman »

I think that there is definitely some use in this thread, but rather than an exercise to try and prove some ranking of the drivers, it should be to discuss the merits of using such a system to rank the drivers.

There are several underlying flaws here in the system, and that is highlighted by the fact that the list is still heavily influenced by the car (Williams drivers at the bottom, top three teams cover the top 5)

The discussion about any simple mathematical model shouldn't be about proving it is right (because a model with this few variables is never going to be give any level of accuracy, let alone to 3 significant figures) - but rather discussing what can we learn from it. What does it tell us?

pokerman
Posts: 35338
Joined: Wed Dec 19, 2012 1:30 pm

Re: Beleriand's Statistical Corner

Post by pokerman »

I like numbers myself and appeciate anyone trying to come up with a rating of drivers using a scientific method but sorry when I look at your results they seem to make little sense, it reminds me a bit of the F1 matrix.
PF1 Pick 10 Competition

2013: 5th Place
2014: Champion
2015: 3rd Place
2016: 4th Place

2017: 9th Place
2018: 7th place

Wins: Canada 2018, Abu Dhabi 2017
Podiums: (8)

Beleriand_K
Posts: 25
Joined: Wed May 20, 2020 2:15 pm

Re: Beleriand's Statistical Corner

Post by Beleriand_K »

Alienturnedhuman wrote:I think that there is definitely some use in this thread, but rather than an exercise to try and prove some ranking of the drivers, it should be to discuss the merits of using such a system to rank the drivers.

There are several underlying flaws here in the system, and that is highlighted by the fact that the list is still heavily influenced by the car (Williams drivers at the bottom, top three teams cover the top 5)

The discussion about any simple mathematical model shouldn't be about proving it is right (because a model with this few variables is never going to be give any level of accuracy, let alone to 3 significant figures) - but rather discussing what can we learn from it. What does it tell us?
It's not surprising that the best drivers drive for the best teams, so I don't see a problem with the drivers from the top 3 teams at top and two of the bottom five places filled by Williams' drivers. As I see it, it is much more interesting, that George Russel driving for the most hopeless team in years still manage to keep four other drivers under him in the ranking. That gives us some important information about his exciting potential. Racing for Williams doesn't automatically demote you to the lowest two position at the list.

But basically I agree with what you say in your post. As I pointed out in the first post, I'm fully aware that there isn't a bulletproof way of separating driver/car performance, even though we all know there is a balance between them that is very important for positions. That's what makes this interesting.

So this thread is open to all creative ideas and point of views about how to rank drivers. I will stick to my version until something better turns up, but I'm definitely willing to listen to a discussion of alternative possibilities.

Herb
Posts: 2352
Joined: Wed Dec 04, 2013 4:42 pm

Re: Beleriand's Statistical Corner

Post by Herb »

Firstly, can I just add that I am always keen to see people conduct rankings like this, so thanks for sharing! I spent a long time in this kind of work, and did an MSc in a related field a few years ago too. As I said above, I'd love to see your formulae.
Beleriand_K wrote:
Herb wrote:Once a driver is in the lead, there is no point in pushing the car more than necessary. They also need to make the car parts last as long as possible, so it is well known that drivers will sometimes take their foot off the gas - leading to a slower average lap time. This may not be possible for the cars in the midfield, that are often in a dogfight for large parts of the race.
I think the problem with having to take the foot of the gas for technical reasons hit everybody once in a while. And the midfield cars are not necessarily under more pressure than the top cars. Kevin Magnussen said once after a race, that it had been very boring, because he was far behind the top 3 teams, but strong enough to keep a safe distance to the rest of the field. So he had just finished lap after lap pretty much on his own. That was once upon a time, when Haas was a fast car...
It does hit everyone once in a while, but I suspect it hits the front drivers more often than the midfield drivers. Hence why I think your formula will punish faster drivers more often. It's kind of the same reason that we have scoffed at fastest laps over the last decade or so - it was often taken by a midfield driver who had something to push for.

Can I also ask how you are handling retirements?

As you brought up Alonso, where would he have featured in your ranking for his last season?

User avatar
Alienturnedhuman
Posts: 3899
Joined: Thu Feb 13, 2003 9:39 pm

Re: Beleriand's Statistical Corner

Post by Alienturnedhuman »

Can you provide some kind of formula?

For example, if we have Team X, with drivers A & B, can you show how you take their laptimes and convert it into a ranking? In terms of variables?

Beleriand_K
Posts: 25
Joined: Wed May 20, 2020 2:15 pm

Re: Beleriand's Statistical Corner

Post by Beleriand_K »

Herb & Alienturnedhuman

I'll try to explain the formulae. Which isn't really a formula, but a number of formulaes.

In every race it takes every driver a certain time to complete the race. From that time and the number of laps I calculate the average time per lap for every driver and for every team. I then calculate the average teamtime per lap, and the specific teams are then ranked according to their average laptime in relation to this average teamtime. It varies from race to race, but pooled for the past 20 races it paints a pretty accurate picture of Mercedes being clearly fastest and Williams being clearly slowest. Not surprising, but it give us some numbers to work with.

It's possible to point out weak points already here, but the intention is not to create the perfect model or nothing at all. The intention is to make something usefull. Knowing that it is not the undisputable truth. Neither is the official standing by the way.

With this teamtime information for the past 20 races I adjust the drivers time per lap for every one of these 20 races by a factor decided by the average strength/weakness of their team. Let me illustrate this with an example:

In the last race of the 2019 season Hamilton won in 1.34.05.715 which is 102.649 seconds per lap. He had a 16.772 seconds gap to Verstappen on 1.34.22.487 which is 102.954 seconds per lap.

But with Mercedes being a considerably faster car than Red Bull (On average 1.32 percent faster than a "neutral" car and Red Bull being 0.78 faster) I adjust the time by 80 percent of the advantage gained by driving a Mercedes and not an average neutral car. The 80 percent is an estimate of the cars part of the difference with the drivers part being the last 20 percent. And the same for Red Bull.

Now Verstappens average time per lap is adjusted upwards to 103.601 seconds (because Red Bull-drivers are also "penalized" for being in a faster car than the average F1 car) and Hamiltons average time per lap is adjusted upwards to 103.732 seconds.

So with the time adjusted for the advantage delivered by a superior car the result now is Verstappen #1 and Hamilton #2. In other words: If Verstappen had been driving the Mercedes and Hamilton the Red Bull, Verstappen would theoretically have won the race by (103.732-103.601) x 55 = 7.205 seconds.

The same calculation is made for all drivers, and that creates a new result of the Abu Dhabi 2019 GP. Sometimes this adjustment causes significant changes to the official result, and other times the adjustment is only minor. But the result of a single race is without much importance.

From the adjusted results from the 20 races in 2019 I calculate a Mean for finishing position, a Standard Deviation and a Standard Error, and with the help of the t-score table I set up a one-sided 90 percent confidence interval. The CI-value is then transformed to a number between 0 and 100 to make the results more readable.

This also answer your (Herb) question about retirements, because a retired driver did not finish the race, and therefore does not figure in the results for that race. In that way a retirement doesn't affect a drivers position neither positively nor negatively directly. But it does have an indirect effect, because fewer completed races creates lower degrees of freedom in the t-score, and therefore lower the drivers total score. Simply because the uncertainty is higher the less data the result is based on.

And now we take a deep breath, and continue to your question about Alonso.

Alonso is a good example of exactly what I just wrote about growing uncertainty with fewer races. I rank Alonso #5 in 2017 (he was #15 in the official standing) and #6 in 2018 (#11 in the official standing), but his results was based on fewer races than most of the other drivers. As you probably remember his McLaren wasn't the most reliable car at that time.

Personably I think his strength would have placed him higher the #5 and #6 if his car had been more reliable. But the important thing here is, that my ranking system placed him 10 and 5 places higher in the final standing than the official standing. Which makes a lot of sense, because he was a world class driver in a not-world class car.
Last edited by Beleriand_K on Sat May 23, 2020 5:23 am, edited 5 times in total.

User avatar
tim3003
Posts: 445
Joined: Mon Mar 13, 2017 11:46 am
Location: Suffolk, UK

Re: Beleriand's Statistical Corner

Post by tim3003 »

Beleriand_K wrote:
The scores are based on a drivers average laptime during a race compared with the average laptime for all the drivers completing the race. This difference is adjusted for the cars average advantage/disadvantage compared with the other teams during the season. During the last 20 races Mercedes preformed on average 1,32% better than a "neutral" team, while Williams on average performed 1,63% worse. It is well known that some cars perform better on some tracks than other, but by including 20 races this should be equalled out.

It is true that two really good drivers can make a car look better than two average drivers, but I believe the law of supply and demand will equal that out, as the best drivers will be snapped up by the best teams, while the average drivers will be stuck with the less attractive teams.
I applaud your efforts, but it seems to me this approach is flawed:

In measuring 'the cars average advantage/disadvantage compared with the other teams during the season' you must include its driver's performance. So the fact that Mercedes were 1.32% better than average may be as much down to Hamilton/Bottas's skills as to the car. This makes the offset for the relative performance of the cars regardless of drivers impossible to judge. Indeed as you say, the best drivers are usually in the best cars, so without Hamilton/Bottas we could conclude the 1.32% advantage for the Mercedes car would be less.

As for a driver's average laptime during a race: It's well-known that a leader will often slow down to protect his car and win by as small a gap as possible, thus increasing his average lap-time. You say only drivers completing the race are included. What about those who don't? Drivers of unreliable cars could be penalised in your system - or even advantaged, it's hard to say. But it's clear there's room for inconsistancy. These days there is also fuel-saving to consider, which will affect some teams more than others - again artificially increasing some drivers' lap-times.

User avatar
Invade
Posts: 3053
Joined: Mon Oct 26, 2015 9:33 pm

Re: Beleriand's Statistical Corner

Post by Invade »

From what I can gather then, this is an attempt to determine a "race pace" ranking rather than a holistic driver ranking. Regarding race pace, I think the pertinent points, pitfalls and problematics have already been outlined and we could discuss ways to better address them.

THE FALLACY OF ‘OBJECTIVE’ SYSTEMS… numbers stripped of context can only ever provide an adumbration and they are maladaptive by design. The obsession with facts, with certain knowledge, fosters an atmosphere in which representation devoid of context is seen as more accurate than fuller sys-tems with more overtly subjective, historical and contextual components. As such, only certainties are permitted, seen as a purification of analysis which is also an incomplete sketch and interpretative framework. The model of F1Metrics first comes to mind in that it measures a sport which is a far cry from a meritocracy, where context is even more vital than in a sport such as tennis, or even those more synchronous in model to production, darts and snooker. To dismiss the subjective component because it is seen as volatile or unreliable and to settle for a half-baked system full of missing parts and then touting it as being far more representative than an opinion drawn through the ‘eye-test’ is disingenuous precisely because it is just as half-baked as that non-analytical judgment, which is, of course, often driven by emotions and biases. There is context between the numbers which can so too be modelled, but always between and embedded in any analytical abstraction is a context which can only be sharpened by communicative means, with an honest and informed rigour. To shy away from this element and yet shower it with lip service upon criticism is not really fooling anyone with a sharp eye. Its repudiation in action is an admittance to a personal view in the pre-eminence of the ‘objective’. This is often a misguided belief. (POLEMIC).

This was an impromptu polemic, I suppose, in reaction to F1Metrics and his attitude toward criticisms of his system taken from an ongoing project. The point I was trying to make is that models typically abandon and saw off a huge proportion of reality when addressing that reality, and yet the allure of steadfast metrics and analytics can sometimes con the system builder into believing they are getting closer to the "truth" through an unfounded skepticism toward the subjective and through a blind belief that "contextless" variables can accurately map out and conjure contextual reality.

I'm just posting this as something to consider and not to attack the OP. I'm actually a fan of systems and models, and trying to find ways to make them as useful as possible

User avatar
DOLOMITE
Posts: 1555
Joined: Sat Mar 06, 2004 8:07 am

Re: Beleriand's Statistical Corner

Post by DOLOMITE »

A good effort Beleriand, very reminiscent of old POBRatings bless 'im.

Anyway, nice work and always good to see some well presented stats.
"I'd rather lose a race going fast enough to win it, than win one going slow enough to lose it".
-Stirling Moss

pokerman
Posts: 35338
Joined: Wed Dec 19, 2012 1:30 pm

Re: Beleriand's Statistical Corner

Post by pokerman »

I think race pace is flawed with engines that have to last 7 races, Hamilton for instance often turns his engine down after his last pit stop in the lead of a race.
PF1 Pick 10 Competition

2013: 5th Place
2014: Champion
2015: 3rd Place
2016: 4th Place

2017: 9th Place
2018: 7th place

Wins: Canada 2018, Abu Dhabi 2017
Podiums: (8)

mikeyg123
Posts: 17831
Joined: Sat Jan 28, 2012 4:13 pm

Re: Beleriand's Statistical Corner

Post by mikeyg123 »

pokerman wrote:I think race pace is flawed with engines that have to last 7 races, Hamilton for instance often turns his engine down after his last pit stop in the lead of a race.
To be fair lots of drivers will do that. What will have more effect is getting lapped. Russell and Kubica will be losing seconds of time on a lot of laps during the race.

pokerman
Posts: 35338
Joined: Wed Dec 19, 2012 1:30 pm

Re: Beleriand's Statistical Corner

Post by pokerman »

mikeyg123 wrote:
pokerman wrote:I think race pace is flawed with engines that have to last 7 races, Hamilton for instance often turns his engine down after his last pit stop in the lead of a race.
To be fair lots of drivers will do that. What will have more effect is getting lapped. Russell and Kubica will be losing seconds of time on a lot of laps during the race.
Yes I was just giving that one example, how can you judge race pace when drivers are not always driving flat out?
PF1 Pick 10 Competition

2013: 5th Place
2014: Champion
2015: 3rd Place
2016: 4th Place

2017: 9th Place
2018: 7th place

Wins: Canada 2018, Abu Dhabi 2017
Podiums: (8)

User avatar
DOLOMITE
Posts: 1555
Joined: Sat Mar 06, 2004 8:07 am

Re: Beleriand's Statistical Corner

Post by DOLOMITE »

I can't be the only one who had to Google "adumbration"....
"I'd rather lose a race going fast enough to win it, than win one going slow enough to lose it".
-Stirling Moss

Beleriand_K
Posts: 25
Joined: Wed May 20, 2020 2:15 pm

Re: Beleriand's Statistical Corner

Post by Beleriand_K »

Three races has been completed, and it is time for an update on the Drivers Rating:
Top 20%Lewis Hamilton62,4
Charles Leclerc61,6
Max Verstappen58,8
Valtteri Bottas56,3
Inside Top 10Sergio Perez55,1
Kimi Raikkönen52,9
Daniel Ricciardo51,9
Carlos Sainz Jr.51,1
Sebastian Vettel50,8
Lando Norris47,1
Outside Top 10Daniil Kvyat45,9
Alex Albon44,9
Kevin Magnussen44,4
Lance Stroll43,5
Antonio Giovinazzi43,1
Romain Grosjean39,3
Bottom 20%George Russell38,4
Pierre Gasly37,9
Nicholas Latifi24,2
Esteban Ocon13,3
Comments
1) Lewis Hamilton has retaken the lead ahead of the two youngsters Leclerc and Verstappen.

2) Sebastian Vettel has dropped out of the top 20%, and is now placed in the midfield. He is now ranked below Sergio Perez, the driver he is likely to replace in Racing Point, when they become Aston Martin next season.

3) Lance Stroll has left the bottom 20%, and has entered the bottom half of the midfield.

4) Nicholas Latifi and Esteban Ocon has completed too few races to justify a realistic score. So their positions should be ignored at this time of the season.

WHoff78
Posts: 778
Joined: Wed Dec 11, 2013 11:01 pm

Re: Beleriand's Statistical Corner

Post by WHoff78 »

An interesting exercise, and always good to see how the numbers fall out, and what patterns might emerge.

Some above have already mentioned some of the perhaps, more significant drawbacks – mainly the fact that the drivers also greatly influence the car, and part of being a great driver, or driver pairing for that matter, is to get the car performing exactly how you like it so you can extract every bit of pace and push it to the next level. Also the management of tyre wear/components and importance of finishing a race as slowly as possible. Which have been acknowledged but I wanted to just raise a couple of other considerations.

The first is how much each team focuses development or optimization of their car towards one specific driver or balances this between the two. Mercedes vs Red Bull being a great example. Enjoying the advantage that they do, one might fully expect the Mercedes team to balance development between both drivers needs as it is actually in their interest to have a close race at the front between their drivers, particularly at circuits where their closest competitors are well of the pace. Teams where both drivers needs are fairly consistent would provide even more benefit as this would allow greater cohesion and efficiency across the team as they develop the car anyway. For Red Bull, it is much more likely that they might focus more of the development and resource around extracting lap time for Max because ultimately he is going to be the one who gets them a couple of race wins over the course of a season. With Merc looking so ominous I expect that would take priority over securing a slightly higher finish for Albon. Of course this is all guess work and no-one can know for sure what approach each team would take.

One other key variable is how much each driver is able to set-up or optimize his car towards overall race pace and maximizing tyre wear / fuel efficiency vs one-lap pace. At certain races some drivers may have the luxury to go more one way that the other. In the midfield there is probably less opportunity for this as giving up a couple of tenths in qualifying could easily put you back a couple of rows on the grid which would offset any gain anyway.

While I actualy think that Bottas is a very solid driver who would hold his own against most on the grid, and would love to see him race directly against someone like Ricciardo or Vettel, I think that both of these considerations could have a big impact on his results. I’m sure they could on most of the drivers though as each team would operate slightly differently and the margins in F1 are pretty small.

And this is not written as an Ocon fan, haha. Not pretty viewing for him though. Appreciate that his score is not relevant at this stage due to number of races.

WHoff78
Posts: 778
Joined: Wed Dec 11, 2013 11:01 pm

Re: Beleriand's Statistical Corner

Post by WHoff78 »

The gap between Kvyat and Gasly is quite interesting - is that influenced by Gasly's stint at Red Bull for the first 12 races of 2019?

Could be interesting to see how the gaps between the Racing point, McLaren and Renault drivers evolve over the season, with drivers getting up to speed in a team and others on their way out.

Beleriand_K
Posts: 25
Joined: Wed May 20, 2020 2:15 pm

Re: Beleriand's Statistical Corner

Post by Beleriand_K »

WHoff78 wrote:The gap between Kvyat and Gasly is quite interesting - is that influenced by Gasly's stint at Red Bull for the first 12 races of 2019?
Yes, I calculate from the last 20 races, so a part of Gasly's Red Bull-period is included in his current ranking. But if we look back to 2018 he didn't impress either with a ranking as #11. That dropped to #17 in the disastrous 2019-season, where he was fired by Red Bull.

Beleriand_K
Posts: 25
Joined: Wed May 20, 2020 2:15 pm

Re: Beleriand's Statistical Corner

Post by Beleriand_K »

The following table shows the highest season ending score during the past three seasons for all drivers completing at least 1/3 of the races in the given season.

It is no surprise that Lewis Hamilton tops the table, while Carlos Sainz, Jr.'s high score in 2017 surprised me e bit. But he did have an excellent 2017-season scoring points in nine of his races for Toro Rosso before his switch from Toro Rosso to Renault. His teammate Daniil Kvyat only scored points in two of the same races.

At the other end of the table Antonio Giovinazzi is the only remaining driver of the last five. Robert Kubica's score shows why his remarkable comeback-attempt only lasted one season. Even though the official table place him above teammate George Russell with one point against zero in 2019, George Russell's score in my table is almost the double of Kubica's.


[/cell][/cell]
Score > 60Lewis Hamilton69.42018
Carlos Sainz Jr.66.52017
Max Verstappen66.12019
Daniel Ricciardo63.52018
Sebastian Vettel63.52017
Charles Leclerc61.52019
Valtteri Bottas60.52017
Fernando Alonso60.52017
Score > 50Sergio Perez59.52017
Nico Hülkenberg56.62018
Kimi Raikkönen56.02017
Felipe Massa55.02017
Esteban Ocon54.02018
Pierre Gasly50.52018
Romain Grosjean50.02017
Score < 50Kevin Magnussen49.02018
Daniil Kvyat48.42019
Lando Norris47.22019
Marcus Ericsson46.52017
Stoffel Vandoorne46.02017
Alex Albon45.42019
Lance Stroll45.32018
Pascal Wehrlein43.52017
George Russell41.52019
Score < 40Antonio Giovinazzi37.62019
Jolyon Palmer37.02017
Brendon Hartley36.42018
Sergey Sirotkin34.42018
Score < 30Robert Kubica23.02019

TheGiantHogweed
Posts: 2814
Joined: Sun Jul 20, 2014 10:15 am

Re: Beleriand's Statistical Corner

Post by TheGiantHogweed »

Beleriand_K wrote:
Sat Aug 01, 2020 9:01 am
The following table shows the highest season ending score during the past three seasons for all drivers completing at least 1/3 of the races in the given season.

It is no surprise that Lewis Hamilton tops the table, while Carlos Sainz, Jr.'s high score in 2017 surprised me e bit. But he did have an excellent 2017-season scoring points in nine of his races for Toro Rosso before his switch from Toro Rosso to Renault. His teammate Daniil Kvyat only scored points in two of the same races.

At the other end of the table Antonio Giovinazzi is the only remaining driver of the last five. Robert Kubica's score shows why his remarkable comeback-attempt only lasted one season. Even though the official table place him above teammate George Russell with one point against zero in 2019, George Russell's score in my table is almost the double of Kubica's.


[/cell][/cell]
Score > 60Lewis Hamilton69.42018
Carlos Sainz Jr.66.52017
Max Verstappen66.12019
Daniel Ricciardo63.52018
Sebastian Vettel63.52017
Charles Leclerc61.52019
Valtteri Bottas60.52017
Fernando Alonso60.52017
Score > 50Sergio Perez59.52017
Nico Hülkenberg56.62018
Kimi Raikkönen56.02017
Felipe Massa55.02017
Esteban Ocon54.02018
Pierre Gasly50.52018
Romain Grosjean50.02017
Score < 50Kevin Magnussen49.02018
Daniil Kvyat48.42019
Lando Norris47.22019
Marcus Ericsson46.52017
Stoffel Vandoorne46.02017
Alex Albon45.42019
Lance Stroll45.32018
Pascal Wehrlein43.52017
George Russell41.52019
Score < 40Antonio Giovinazzi37.62019
Jolyon Palmer37.02017
Brendon Hartley36.42018
Sergey Sirotkin34.42018
Score < 30Robert Kubica23.02019
I admit have been quite critical of the level of praise for Sainz recently as I think he is decent, but not amazing. I think Norris matching him so far is proving that. But 2017 was not a good season for him. Possibly his worst. He crashed out 3 times and retired and on two of those occasions, collected and ruined another drivers race. He ended his career at Toro rosso by crashing out in the first corner in japan. Not good. Caused more retirements than any other driver that year. He had plenty of reasonable races, but he was against Kvyat who was massively underperforming (probably rated as one of the worst drivers on the grid at the time). The 2nd highest score covering every drivers season over the past 3 just seems incredibly generous. I would question weather he could be in the top 10 rankings that season alone.

Beleriand_K
Posts: 25
Joined: Wed May 20, 2020 2:15 pm

Re: Beleriand's Statistical Corner

Post by Beleriand_K »

TheGiantHogweed wrote:
Sat Aug 01, 2020 9:28 am

I admit have been quite critical of the level of praise for Sainz recently as I think he is decent, but not amazing. I think Norris matching him so far is proving that. But 2017 was not a good season for him. Possibly his worst. He crashed out 3 times and retired and on two of those occasions, collected and ruined another drivers race. He ended his career at Toro rosso by crashing out in the first corner in japan. Not good. Caused more retirements than any other driver that year. He had plenty of reasonable races, but he was against Kvyat who was massively underperforming (probably rated as one of the worst drivers on the grid at the time). The 2nd highest score covering every drivers season over the past 3 just seems incredibly generous. I would question weather he could be in the top 10 rankings that season alone.
I tend to agree with you about the star quality of Carlos Sainz Jr., but Ferrari can't be that wrong...?

He's ranked #14 and #9 in 2018 and 2019, but I don't agree with you about 2017 being his worst season. It's true, that he did crash out a number of times, but his average position when completing a race was 7,4. That was not only considerably better than Kvyats 12,2, it was also achieved in a Toro Rosso with the same Renault-engine that made Red Bulls Christian Horner loose his temper several times that season.

Beleriand_K
Posts: 25
Joined: Wed May 20, 2020 2:15 pm

Re: Beleriand's Statistical Corner

Post by Beleriand_K »

Sebastian Vettels dream of replicating the Ferrari-performance of his childhood hero Michael Schumacher didn't come through. After more or less being fired by Ferrari, Sebastian Vettels last season with the team is turning into a farce.

He will probably consider his time with Ferrari a failure, as he didn't even come close to what Michael Schumacer achieved, but is that a fair judgement?

I don't think so. First of all, Michael Schumachers results with Ferrari was of another world. And secondly, Sebastian Vettel actually didn't do that bad with Ferrari, if we look at his results against a long list of Ferrari-drivers throughout history. This list includes drivers with more than ten races for Ferrari, and Sebastian Vettel is the fourth most successful Ferrari-driver ever. Not bad at all after having won four WDC with another team first.

1Michael Schumacher2878
Niki Lauda1723
Alberto Ascari1713
Sebastian Vettel1540
Fernando Alonso1358
Rubens Barrichello1348
Alain Prost982
Kimi Räikkönen975
Mike Hawthorn941
Felipe Massa901
11Phil Hill867
John Surtees834
Carlos Reutemann810
Jacky Ickx780
Eddie Irvine780
Giuseppe Farina763
Charles Leclerc737
Gerhard Berger695
Nigel Mansell691
René Arnoux682
21Clay Regazzoni670
Gilles Villeneuve650
José Froilán González629
Peter Collins618
Michele Alboreto588
Patrick Tambay573
Jody Scheckter562
Wolfgang von Trips480
Jean Alesi460
Didier Pironi438
31Lorenzo Bandini400
Luigi Villoresi329
Luigi Musso312
Maurice Trintignant301
Stefan Johansson292
Chris Amon233
Piero Taruffi199
Mario Andretti120
Eugenio Castellotti35
Arturo Merzario10
41Ivan Capelli9
Last edited by Beleriand_K on Mon Aug 10, 2020 2:36 pm, edited 3 times in total.

Asphalt_World
Posts: 4953
Joined: Thu May 12, 2011 6:08 pm

Re: Beleriand's Statistical Corner

Post by Asphalt_World »

What are the numbers?
Instagram @simply_italian_cars

j man
Posts: 3494
Joined: Sun Aug 28, 2011 8:48 pm
Location: UK

Re: Beleriand's Statistical Corner

Post by j man »

Subjectively speaking, I don't see how Vettel's Ferrari career can be viewed as a success given that he had the car to deliver at least 1 championship and he didn't, and I don't see how he can be ranked above the drivers who did. Or ranked above Alonso who never really had the car to deliver a championship but very nearly did anyway, on two occasions.

Beleriand_K
Posts: 25
Joined: Wed May 20, 2020 2:15 pm

Re: Beleriand's Statistical Corner

Post by Beleriand_K »

Asphalt_World wrote:
Sun Aug 09, 2020 4:14 pm
What are the numbers?
It's a calculated score based on achieved results.

Paolo_Lasardi
Posts: 2570
Joined: Sat May 26, 2012 2:04 pm

Re: Beleriand's Statistical Corner

Post by Paolo_Lasardi »

Beleriand_K wrote:
Sun Aug 09, 2020 6:01 pm
Asphalt_World wrote:
Sun Aug 09, 2020 4:14 pm
What are the numbers?
It's a calculated score based on achieved results.
A bit more precise?

A "score based on results" could be anything . ..

Asphalt_World
Posts: 4953
Joined: Thu May 12, 2011 6:08 pm

Re: Beleriand's Statistical Corner

Post by Asphalt_World »

Beleriand_K wrote:
Sun Aug 09, 2020 6:01 pm
Asphalt_World wrote:
Sun Aug 09, 2020 4:14 pm
What are the numbers?
It's a calculated score based on achieved results.
How is it calculated, though?
Instagram @simply_italian_cars

Beleriand_K
Posts: 25
Joined: Wed May 20, 2020 2:15 pm

Re: Beleriand's Statistical Corner

Post by Beleriand_K »

Asphalt_World wrote:
Sun Aug 09, 2020 7:15 pm
Beleriand_K wrote:
Sun Aug 09, 2020 6:01 pm
Asphalt_World wrote:
Sun Aug 09, 2020 4:14 pm
What are the numbers?
It's a calculated score based on achieved results.
How is it calculated, though?
The score is the result of this Excel-formula:

=(H3+I3/2+J3/4+K3/8+L3/16+M3/32)/N3*(1-1/LOG(N3))*10000-ROW(A3)/10000

H is total wins, I is total seconds and so on. This formula has already been discussed intensely here on PlanetF1 a few years ago, so the answer to all questions can be found by searching in PlanetF1's older threads.

Paolo_Lasardi
Posts: 2570
Joined: Sat May 26, 2012 2:04 pm

Re: Beleriand's Statistical Corner

Post by Paolo_Lasardi »

Beleriand_K wrote:
Mon Aug 10, 2020 6:06 am
Asphalt_World wrote:
Sun Aug 09, 2020 7:15 pm
Beleriand_K wrote:
Sun Aug 09, 2020 6:01 pm
Asphalt_World wrote:
Sun Aug 09, 2020 4:14 pm
What are the numbers?
It's a calculated score based on achieved results.
How is it calculated, though?
The score is the result of this Excel-formula:

=(H3+I3/2+J3/4+K3/8+L3/16+M3/32)/N3*(1-1/LOG(N3))*10000-ROW(A3)/10000

H is total wins, I is total seconds and so on. This formula has already been discussed intensely here on PlanetF1 a few years ago, so the answer to all questions can be found by searching in PlanetF1's older threads.
What is N and what is A?

Beleriand_K
Posts: 25
Joined: Wed May 20, 2020 2:15 pm

Re: Beleriand's Statistical Corner

Post by Beleriand_K »

Paolo_Lasardi wrote:
Mon Aug 10, 2020 8:04 am
Beleriand_K wrote:
Mon Aug 10, 2020 6:06 am

The score is the result of this Excel-formula:

=(H3+I3/2+J3/4+K3/8+L3/16+M3/32)/N3*(1-1/LOG(N3))*10000-ROW(A3)/10000

H is total wins, I is total seconds and so on. This formula has already been discussed intensely here on PlanetF1 a few years ago, so the answer to all questions can be found by searching in PlanetF1's older threads.
What is N and what is A?
N is total number of races and A is just the A-column in Excel. "-ROW(A3)/10000" doesn't affect the score, but makes sure, that Excel doesn't have problems sorting the lines by score, because two scores are identical.

pokerman
Posts: 35338
Joined: Wed Dec 19, 2012 1:30 pm

Re: Beleriand's Statistical Corner

Post by pokerman »

Beleriand_K wrote:
Sun Aug 09, 2020 4:13 pm
Sebastian Vettels dream of replicating the Ferrari-performance of his childhood hero Michael Schumacher didn't come through. After more or less being fired by Ferrari, Sebastian Vettels last season with the team is turning into a farce.

He will probably consider his time with Ferrari a failure, as he didn't even come close to what Michael Schumacer achieved, but is that a fair judgement?

I don't think so. First of all, Michael Schumachers results with Ferrari was of another world. And secondly, Sebastian Vettel actually didn't do that bad with Ferrari, if we look at his results against a long list of Ferrari-drivers throughout history. This list includes drivers with more than ten races for Ferrari, and Sebastian Vettel is the fourth most successful Ferrari-driver ever. Not bad at all after having won four WDC with another team first.

1Michael Schumacher2878
Niki Lauda1723
Alberto Ascari1713
Sebastian Vettel1492
Fernando Alonso1358
Rubens Barrichello1348
Alain Prost982
Kimi Räikkönen975
Mike Hawthorn941
Felipe Massa901
11Phil Hill867
John Surtees834
Carlos Reutemann810
Jacky Ickx780
Eddie Irvine780
Giuseppe Farina763
Charles Leclerc737
Gerhard Berger695
Nigel Mansell691
René Arnoux682
21Clay Regazzoni670
Gilles Villeneuve650
José Froilán González629
Peter Collins618
Michele Alboreto588
Patrick Tambay573
Jody Scheckter562
Wolfgang von Trips480
Jean Alesi460
Didier Pironi438
31Lorenzo Bandini400
Luigi Villoresi329
Luigi Musso312
Maurice Trintignant301
Stefan Johansson292
Chris Amon233
Piero Taruffi199
Mario Andretti120
Eugenio Castellotti35
Arturo Merzario10
41Ivan Capelli9
In respect to Alonso, Vettel was given better cars.
PF1 Pick 10 Competition

2013: 5th Place
2014: Champion
2015: 3rd Place
2016: 4th Place

2017: 9th Place
2018: 7th place

Wins: Canada 2018, Abu Dhabi 2017
Podiums: (8)

Herb
Posts: 2352
Joined: Wed Dec 04, 2013 4:42 pm

Re: Beleriand's Statistical Corner

Post by Herb »

Beleriand_K wrote:
Mon Aug 10, 2020 6:06 am
Asphalt_World wrote:
Sun Aug 09, 2020 7:15 pm
Beleriand_K wrote:
Sun Aug 09, 2020 6:01 pm
Asphalt_World wrote:
Sun Aug 09, 2020 4:14 pm
What are the numbers?
It's a calculated score based on achieved results.
How is it calculated, though?
The score is the result of this Excel-formula:

=(H3+I3/2+J3/4+K3/8+L3/16+M3/32)/N3*(1-1/LOG(N3))*10000-ROW(A3)/10000

H is total wins, I is total seconds and so on. This formula has already been discussed intensely here on PlanetF1 a few years ago, so the answer to all questions can be found by searching in PlanetF1's older threads.
Because this stuff interests me, I've been trying to replicate your results. Using your formula, I get 1529 for Vettel* - so I must be missing something, but anything I can think of doing (like excluding a retirement/DNS), would only increase that score.

*using wikipedia's season summary of his years in Ferrari - which I haven't checked!

Post Reply