Member
Posted 21 November 2022 - 08:26
ALO’s H2H ratio is impressive considering that a) he is in his 40s and b) Ocon is very solid in Q. Remember how he trounced Perez couple of years ago.
I don't know. Ocon trounced Perez indeed but on the other hand was trounced by Ricciardo. I'm not completely sure he is that fast on one lap pace.
Advertisement
Member
Posted 21 November 2022 - 08:50
Do we?
Ricciardo was known as the best qualifier on the grid until Max started beating him. Sainz was no slouch either but was easily beaten by Max in their second season together, until Max moved on to Red Bull.
Maybe Lewis’ qualifying skills are just a myth?
Until 2 drivers are up against each other in the same car, no one can really tell who has the upper hand.
Member
Posted 21 November 2022 - 11:08
Member
Posted 21 November 2022 - 11:20
Fair play to Sainz actually. He's much closer on average to Leclerc (One of the best, if not THE best at one lap) than I thought.
Member
Posted 21 November 2022 - 12:07
Can I just say to all who don't appreciate Tyker's choice of analysis - you can always make your own and post it in the thread. If you don't agree with how he reviews it, it isn't hard to run the numbers yourself and create a table as a counter argument for how you personally think is the fairest way to analyse the qualifying duels.
Clicking "Like" for this post is not enough, we need to blast trumpets and wave flags too. Tyker has come up with a rational method of comparing, and has explained the logic behind it. Nobody's claiming it's the only possible way, but it's a pretty good one and gives information some of us regard as extremely interesting.
Member
Posted 21 November 2022 - 14:09
Can I just say to all who don't appreciate Tyker's choice of analysis - you can always make your own and post it in the thread. If you don't agree with how he reviews it, it isn't hard to run the numbers yourself and create a table as a counter argument for how you personally think is the fairest way to analyse the qualifying duels.
I'll present my alternative. My methodology is:
- Take times from the last session both drivers set a flying lap in (to mitigate against changing conditions between Q1/Q2/Q3).
- The average laptime in 2022 was 87s, so all gaps are equalised to an 87s laptime, to try to make the gaps consistent when looking at, e.g., a dry Red Bull Ring vs a wet Singapore.
- The top and bottom 3 gaps between teammates are disregarded, to get a more median average. This should eliminate freak results / bad luck from the numbers. Thus the below is from the closest 16 races for each pairing.
Using this methodology, the teammate gaps were:
1. ALB > LAT = -0.497
2. VER > PER = -0.475
3. NOR > RIC = -0.384
4. MAG > SCH = -0.292
5. ALO > OCO = -0.264
6. BOT > ZHO = -0.250
7. GAS > TSU = -0.199
8. VET > STR = -0.152
9. HAM > RUS = -0.116
10. LEC > SAI = -0.046
The takeaways are...
- Perez is as abject in qualifying against Verstappen as Latifi is against Albon. It also means he is the outlier in that he is the only one of the top 4 worst performing qualifiers vs teammates to keep his drive for 2023.
- Leclerc is not the dominant qualifier over Sainz that many would assume, and were the closest two teammates.
- The Alpha Tauris and Aston Martins were pretty close, some might say surprisingly.
- Alonso and Hamilton aren't bad for old men.
Edited by JimmyClark, 21 November 2022 - 14:18.
New Member
Posted 21 November 2022 - 14:50
Maybe not a myth, but certainly over blown by some. He's never really destroyed any teammate that I can remember, apart from Button, who has a pretty poor record overall in qualifying.
He's definitely a quick qualifier but his pace over 1 lap has never been the thing that separated him from his team mates. As you say, Button is the only one he consistently beat by a large margin and he was generally quite a poor qualifier throughout his career. Button was totally outpaced by drivers like Ralf, Fisi, Trulli and even lost out to Barrichello and Perez in quali so Lewis really should have absolutely destroyed him. So despite the crazy number of poles, I'd say Hamilton is more in the Alonso mould where it's his racecraft that is all time top tier rather than his one lap pace
Member
Posted 21 November 2022 - 16:09
I'll present my alternative. My methodology is:
- Take times from the last session both drivers set a flying lap in (to mitigate against changing conditions between Q1/Q2/Q3).
- The average laptime in 2022 was 87s, so all gaps are equalised to an 87s laptime, to try to make the gaps consistent when looking at, e.g., a dry Red Bull Ring vs a wet Singapore.
- The top and bottom 3 gaps between teammates are disregarded, to get a more median average. This should eliminate freak results / bad luck from the numbers. Thus the below is from the closest 16 races for each pairing.
Using this methodology, the teammate gaps were:
1. ALB > LAT = -0.497
2. VER > PER = -0.475
3. NOR > RIC = -0.384
4. MAG > SCH = -0.292
5. ALO > OCO = -0.264
6. BOT > ZHO = -0.250
7. GAS > TSU = -0.199
8. VET > STR = -0.152
9. HAM > RUS = -0.116
10. LEC > SAI = -0.046
The takeaways are...
- Perez is as abject in qualifying against Verstappen as Latifi is against Albon. It also means he is the outlier in that he is the only one of the top 4 worst performing qualifiers vs teammates to keep his drive for 2023.
- Leclerc is not the dominant qualifier over Sainz that many would assume, and were the closest two teammates.
- The Alpha Tauris and Aston Martins were pretty close, some might say surprisingly.
- Alonso and Hamilton aren't bad for old men.
Thanks for this very thoughtful approach to analysis of the numbers, James.
New Member
Posted 21 November 2022 - 16:39
I'll present my alternative. My methodology is:
- Take times from the last session both drivers set a flying lap in (to mitigate against changing conditions between Q1/Q2/Q3).
- The average laptime in 2022 was 87s, so all gaps are equalised to an 87s laptime, to try to make the gaps consistent when looking at, e.g., a dry Red Bull Ring vs a wet Singapore.
- The top and bottom 3 gaps between teammates are disregarded, to get a more median average. This should eliminate freak results / bad luck from the numbers. Thus the below is from the closest 16 races for each pairing.
Using this methodology, the teammate gaps were:
1. ALB > LAT = -0.497
2. VER > PER = -0.475
3. NOR > RIC = -0.384
4. MAG > SCH = -0.292
5. ALO > OCO = -0.264
6. BOT > ZHO = -0.250
7. GAS > TSU = -0.199
8. VET > STR = -0.152
9. HAM > RUS = -0.116
10. LEC > SAI = -0.046
The takeaways are...
- Perez is as abject in qualifying against Verstappen as Latifi is against Albon. It also means he is the outlier in that he is the only one of the top 4 worst performing qualifiers vs teammates to keep his drive for 2023.
- Leclerc is not the dominant qualifier over Sainz that many would assume, and were the closest two teammates.
- The Alpha Tauris and Aston Martins were pretty close, some might say surprisingly.
- Alonso and Hamilton aren't bad for old men.
This does feel like a more accurate representation of the differences in quali pace this season to me. Aside from Leclerc vs Sainz (which I felt was closer to 2 tenths on average) and Verstappen to Perez (felt closer to 3 tenths to me), I'd say that this was almost perfectly in line with what I'd expect the difference to be between the drivers on any given weekend.
Edited by Ben24, 21 November 2022 - 16:43.
Member
Posted 21 November 2022 - 18:19
This does feel like a more accurate representation of the differences in quali pace this season to me. Aside from Leclerc vs Sainz (which I felt was closer to 2 tenths on average) and Verstappen to Perez (felt closer to 3 tenths to me), I'd say that this was almost perfectly in line with what I'd expect the difference to be between the drivers on any given weekend.
I ran the numbers too for the rb-pairing as I also felt it to be smaller, and I got per being slower by 0.297s(, and 0.247s if you don't omit the three smallest gaps). I stopped there.
Member
Posted 21 November 2022 - 21:07
I ran the numbers too for the rb-pairing as I also felt it to be smaller, and I got per being slower by 0.297s(, and 0.247s if you don't omit the three smallest gaps). I stopped there.
Member
Posted 21 November 2022 - 23:09
Did you...
A) only count times set in the same session?
B) Equalise to 87 seconds?
C) discount both top three and bottom three most extreme times?
I'll see if I can share my spreadsheet for transparency when I get home.
No system for working this out will be perfect, but I tried to make one that is fairest to show relative performance for much of the time between two drivers.
wrt A+B+C, yes, I did.
Member
Posted 22 November 2022 - 16:01
wrt A+B+C, yes, I did.
Edited by JimmyClark, 22 November 2022 - 21:15.
Member
Posted 22 November 2022 - 19:43
Yes, I see what you did there, and I have to say, I already suspected it to be the case when I saw the methodology.
What you did in your execution, is different from what you said about the 3rd rule:
I'll present my alternative. My methodology is:
- Take times from the last session both drivers set a flying lap in (to mitigate against changing conditions between Q1/Q2/Q3).
- The average laptime in 2022 was 87s, so all gaps are equalised to an 87s laptime, to try to make the gaps consistent when looking at, e.g., a dry Red Bull Ring vs a wet Singapore.
- The top and bottom 3 gaps between teammates are disregarded, to get a more median average. This should eliminate freak results / bad luck from the numbers. Thus the below is from the closest 16 races for each pairing.
What you did in your calculations was disregard the top-3 quali-wins (delta's measured in time) from each driver. But that's not what your 3rd rule says. It says to disregard the 3 biggest and 3 smallest gaps (again, measured in time). That's what I did, so I excluded R5 and R7, whereas you excluded R2 and R17. Pretty big difference, and to pass judgment between the two, I'd say the way how it was originally written down by you,.is a better way than to take away the 3 biggest wins from each driver per definition.
That's by far the biggest thing, but another is that in Austria you took their Q2-time while I took their Q3-time, because that corresponds with your 1st rule (this was the instance they deleted PER's fastest lap in Q2 because of track limits.. after having participated in Q3). Over the 16 races that need to be accounted for, this explains about 0.05s, so not that much.
Lastly, I see you have a delta of 1.045s in Miami, but the delta was 0.045s. However, seeing the 0.044-figure in brackets, I guess you still used the right delta.
This is just from first glance, could be more differences (haven't checked the delta's(, aside from aforementioned Miami and Aut, and the one from R1 Bahrain), and calculations).
Member
Posted 22 November 2022 - 20:05
Member
Posted 22 November 2022 - 21:14
Oh that's a misunderstanding then, apologies if it wasn't clear. If you have the spectrum of teammate differences, then doing an excel formula to show top 3 / bottom 3 takes out the six extremes. If you do it for when one driver is in front of the other, that unfairly skews it against the better performing driver.Yes, I see what you did there, and I have to say, I already suspected it to be the case when I saw the methodology.
What you did in your execution, is different from what you said about the 3rd rule:
What you did in your calculations was disregard the top-3 quali-wins (delta's measured in time) from each driver. But that's not what your 3rd rule says. It says to disregard the 3 biggest and 3 smallest gaps (again, measured in time). That's what I did, so I excluded R5 and R7, whereas you excluded R2 and R17. Pretty big difference, and to pass judgment between the two, I'd say the way how it was originally written down by you,.is a better way than to take away the 3 biggest wins from each driver per definition.
I went off the official qualifying standings, where both set a competitive time in similar conditions to try not to skew it too much (and deleted laps would not have counted). But the top 3/bottom 3 should help eliminate extremes having too much influence, which is the point of this exercise. I would have thought my Brazil choice would be more controversial to be honest...That's by far the biggest thing, but another is that in Austria you took their Q2-time while I took their Q3-time, because that corresponds with your 1st rule (this was the instance they deleted PER's fastest lap in Q2 because of track limits.. after having participated in Q3). Over the 16 races that need to be accounted for, this explains about 0.05s, so not that much.
Oh that's a typo in this post; it was correct in the spreadsheet. I'll amend.Lastly, I see you have a delta of 1.045s in Miami, but the delta was 0.045s. However, seeing the 0.044-figure in brackets, I guess you still used the right delta.
I think there's just a fundamental difference between how we define top/bottom 3. But as the post above says, if you disagree then present your own data. I think mine is the fairest I can think of given the variables of comparing qualifying performance. But happy to be contested on that.This is just from first glance, could be more differences (haven't checked the delta's(, aside from aforementioned Miami and Aut, and the one from R1 Bahrain), and calculations).
It's exactly the same proportionally - if the average laptime of the year was 100s then I would have used that. I just presented the calculation that way so that a more relatable number was given when comparing laptimes. Though to be honest, numerically, the difference really isn't that much.Why don't you use % instead of the 87s lap time? You can calculate the time depending on the laptime then.
Edited by JimmyClark, 22 November 2022 - 21:44.
Member
Posted 22 November 2022 - 22:58
Oh that's a misunderstanding then, apologies if it wasn't clear. If you have the spectrum of teammate differences, then doing an excel formula to show top 3 / bottom 3 takes out the six extremes. If you do it for when one driver is in front of the other, that unfairly skews it against the better performing driver.
No, it's not. What you get by what you did, taking the best (relative to the teammate, measured in time) three quali-performances out of the equation for each driver, is taking out almost any counterweight PER had. They represent 75% of his quali-wins. Furthermore, 2 of those 3 were wins on merit and no
freak results / bad luck
whatsoever; they were wins with a delta to his teammate of one third of a second, combined. Disregarding these serve as a double whammy on top.
Your goal was to eliminate those freak results/bad luck; well, outliers don't discriminate. So if one driver has all three of them in his favour, it's perfecly fine to eliminate only those instead of deleting 2 normal quali-wins of the other driver on top, if your goal is to not skew 'true performance analysis'.
I went off the official qualifying standings, where both set a competitive time in similar conditions to try not to skew it too much (and deleted laps would not have counted). But the top 3/bottom 3 should help eliminate extremes having too much influence, which is the point of this exercise. I would have thought my Brazil choice would be more controversial to be honest...
Yes I know you did that, but I just said that I simply adhered to rule nr 1:
- Take times from the last session both drivers set a flying lap in (to mitigate against changing conditions between Q1/Q2/Q3).
All in all, 'Austria' doesn't change much, about 0.05s as previously stated; the eliminating process doesn't have much to do with Austria in the first place, as we both counted it.
Brazil? Why would thát be controversial, or are you suggesting the gap of 4.638s over an 87s lap is the true performance gap between the two, considering "eliminate extremes having too much influence, which is the point of this exercise"?
I think there's just a fundamental difference between how we define top/bottom 3. But as the post above says, if you disagree then present your own data. I think mine is the fairest I can think of given the variables of comparing qualifying performance. But happy to be contested on that.
Well I think it's more a case of a misunderstanding, say a communication thing. I would have worded it differently if my methodology entailed to disregard the best three quali's of each driver, instead of speaking about
- The top and bottom 3 gaps between teammates
, no biggie, that's why we can reply to one another. 
About me presenting my own idea, I'll pass on that, for now. This just started by me having the same feeling as Ben24, namely that the fer and rb intra-team gaps were bigger/smaller, and me wanting to verify that, but still using the system in your post. So I did (for rb), that's all.
But yeah, as I said before in this post, I don't think it's fair to disregard the 3 biggest wins of a driver per definition, because it can cause the figure to skew a lot, as is the case now, where PER has 2 of his 3 'wins on merit' deleted, while the only one that should be discounted for such an exercise IMO is his 4th win in SIN.
Edited by eab, 22 November 2022 - 23:00.
Member
Posted 23 November 2022 - 01:59
Sainz was no slouch either but was easily beaten by Max in their second season together, until Max moved on to Red Bull.
We're strictly talking about qualifying here, so this is really disingenous. "Easily beaten by Max in their second season together". A season that lasted a grand total of 4 races. Max did qualify better in 3 of those 4, but the average gap would be neck and neck since Max beat him by a tenth in those 3 ocassions (not even that in Bahrain, as he was ahead only by 0.044) while Sainz outqualified Max in China for over 3 tenths. So the average is more or less a tie, while in 2015 Sainz was the one ahead? How is that "easily beaten"?
Edited by GunnarN7, 23 November 2022 - 02:01.
Member
Posted 23 November 2022 - 10:47
No, it's not. What you get by what you did, taking the best (relative to the teammate, measured in time) three quali-performances out of the equation for each driver, is taking out almost any counterweight PER had. They represent 75% of his quali-wins. Furthermore, 2 of those 3 were wins on merit and no
whatsoever; they were wins with a delta to his teammate of one third of a second, combined. Disregarding these serve as a double whammy on top.
Your goal was to eliminate those freak results/bad luck; well, outliers don't discriminate. So if one driver has all three of them in his favour, it's perfecly fine to eliminate only those instead of deleting 2 normal quali-wins of the other driver on top, if your goal is to not skew 'true performance analysis'.
But yeah, as I said before in this post, I don't think it's fair to disregard the 3 biggest wins of a driver per definition, because it can cause the figure to skew a lot, as is the case now, where PER has 2 of his 3 'wins on merit' deleted, while the only one that should be discounted for such an exercise IMO is his 4th win in SIN.
But, statistically, a weekend where a driver 'wins' doesn't matter when you are finding an overall performance trend. If it was split 50/50 throughout the season, then yes maybe. Eliminating 75% of Perez's 'wins' is fine because 82% of the time Verstappen was faster anyway.
To put it another way, think of a hypothetical 22-race season where each track is equal in length and Driver A qualifies 0.1s ahead of Driver B each race weekend, but on one occasion Driver A gets his setup completely wrong, the engine misfires too, and the team puts on the wrong tyres and he qualifies 3s behind Driver B. Any reasonable person would still say, overall for the season, that Driver A is 0.1s faster than Driver B.
If you were doing a straight average of that season, which seems to be the normal methodology for these kind of threads, then it would conclude that Driver B is 0.04s faster than Driver A! This is clearly wrong.
With my method, I'm looking at an average of the median 75% of races (well 73%, but that's roughly it - 75% seems a reasonable statistical sample to take), and it will conclude the average qualifying gap to be that Driver A is 0.1s faster than Driver B, as that extreme anomaly is eliminated. This is exactly in line with what you would expect.
With your variation (as I understand it) would entail keeping Driver B's result, but eliminating six of Driver A's, which would mean the seasonal gap to be +0.09s in favour of Driver B. Again, this is really not the competitve reality between the two.
Obviously this is a very simple example, but you can see from my results that most would agree that these are the respective gaps you expect between drivers pitted against each other on an average circuit in average conditions.
Anyway, this was all meant to be a bit of fun.
Edited by JimmyClark, 23 November 2022 - 10:49.
Advertisement
Member
Posted 23 November 2022 - 13:21
3 meaning Gasly, Albon and Perez. None are renowned for their 1 lap speedDo we?
Ricciardo was known as the best qualifier on the grid until Max started beating him. Sainz was no slouch either but was easily beaten by Max in their second season together, until Max moved on to Red Bull.
Maybe Lewis’ qualifying skills are just a myth?
Until 2 drivers are up against each other in the same car, no one can really tell who has the upper hand.
Edited by lewislorenzo, 23 November 2022 - 13:24.
Member
Posted 23 November 2022 - 14:35
With your variation (as I understand it) would entail keeping Driver B's result, but eliminating six of Driver A's, which would mean the seasonal gap to be +0.09s in favour of Driver B. Again, this is really not the competitve reality between the two.
First of all, there is no 'variation of mine' that I laid out here. I'm repeating myself when I say that what I did was verify the presented results by using your own method the way it was written down by you, the way I understood it, in particular the 3rd rule (as that's what's mainly causing the 92% difference in delta):
I'll present my alternative. My methodology is:
-
-
- The top and bottom 3 gaps between teammates are disregarded
By elaborating further, it came to know that you didn't delete the 3 smallest and 3 biggest gaps, but that you deleted the 3 biggest gap-wins from each driver.
Secondly, no, using the 'delete the 3 biggest gaps' method (overall gaps, not 3 from each driver) I used to come to the 0.247s figure would certainly not mean disregarding 6 wins of driver A and keep the one of B. I'm actually amazed to read how you went from 3 to 6. Nowhere have I used '6', always talking about "3 biggest gaps", and in fact it is you who deletes 6 results, 3 of each driver. Anyway no, it would mean that B's win would be deleted, and 2 of A's. Like I said, outliers do not discriminate. So the result would be driver A wins with an avg of 0.1s per quali.
It turns out that in this instance, you get the same results as when you calculate it the way you used to come to this years' gaps. However, that's mainly because of choosing a convenient scenario, but it actually serves, to an extent, as an example of pointing at a flaw in your methodology which I implied earlier. Your method says: "Delete the 3 biggest wins of A, and the 3 biggest wins of B." However, driver B doesn't even have 3 wins, so why cement that as some kind of an axioma into a method? The way I calculated is superior because it doesn't discriminate with regard to outliers between drivers. If all the outliers happened to one driver, then so be it.
This brings me to the next point. If I were to want to paint me a representative picture of quali delta's, I would not set a limit of disregarding results at all, no minimum, no maximum. The fact that in the rb case my 0.247s figure is still representative by disregarding solely the 3 biggest gaps, is because the only two other quali's I'd disregard were quite evenly big and divided equally amongst the two dirvers. And again, if there were less than 3 anomalies, then I would delete less accordingly.
With my method, I'm looking at an average of the median 75% of races (well 73%, but that's roughly it - 75% seems a reasonable statistical sample to take), and it will conclude the average qualifying gap to be that Driver A is 0.1s faster than Driver B, as that extreme anomaly is eliminated. This is exactly in line with what you would expect.
No, you're wrong. You just chose a convenient example, as I said above. You can get the exact same result by using a lot of different methods, including the method that was used by me to come to the 0.247s figure as I showed above.
But, statistically, a weekend where a driver 'wins' doesn't matter when you are finding an overall performance trend. If it was split 50/50 throughout the season, then yes maybe. Eliminating 75% of Perez's 'wins' is fine because 82% of the time Verstappen was faster anyway.
These again, are erroneous statements. First off, max wasn't "82% of the time faster anyway", 4 of his 18 wins came precisely by those anomalies. Secondly, eliminating 75% of PER's wins is not "fine" at all, as 75% of his wins were on merit. I do not understand why somebody who says his aim is to eliminate fluke results and retain the 'legit' ones to come to a 'true picture', lets the fluke wins of one driver stand, while deleting the wins on merit of the other on top. A double whammy.
Regarding the "a weekend where a driver wins doesn't matter in case of a trend", it does matter. A lot. Just look at this real life example in which your gap is 92% bigger than if you wouldn't have disregarded PER's wins on merit.
And speaking about spotting a trend, when I looked up all these times, it once again became very clear how the development of the car went completely away from PER during the season. From Silverstone onwards it went downhill, and even more so after the summerbreak. Out of the first 7 quali wins on merit, 3 went to PER, so almost even. Out of the rest (10), he won none.
Anyway, this was all meant to be a bit of fun.
Yeah, so I thought, But are you sure? Because you do seem to care a lot about this all, especially about whether the gap between the rb-pair is perceived to be big enough.
Member
Posted 23 November 2022 - 14:51
We're strictly talking about qualifying here, so this is really disingenous. "Easily beaten by Max in their second season together". A season that lasted a grand total of 4 races. Max did qualify better in 3 of those 4, but the average gap would be neck and neck since Max beat him by a tenth in those 3 ocassions (not even that in Bahrain, as he was ahead only by 0.044) while Sainz outqualified Max in China for over 3 tenths. So the average is more or less a tie, while in 2015 Sainz was the one ahead? How is that "easily beaten"?
And there are at least two more caveats. Firstly, two of max' marginal wins came under that fiasco of a quali elimination format that was criticized by everybody because it didn't give everybody a chance to improve on their lap time, and it was thus subsequently dropped at the 3rd GP (plan was to drop it right away but, as seen before and since, certain regulators didn't want to admit defeat right away and so prolonged the format for one more weekend). Secondly, the outright fastest time between the two in quali of race 1 was actually set by SAI, in q2.
So yeah, very, very close those two.
Edited by eab, 23 November 2022 - 14:56.
Member
Posted 23 November 2022 - 14:55
Another poor example, Button...
It is well known that Perez is a very poor qualifier.
RC Forum Host
Posted 23 November 2022 - 15:01
Thanks Tyker for theses stats. I enjoyed them 
.
Member
Posted 23 November 2022 - 15:08
Oh, and about this
Obviously this is a very simple example, but you can see from my results that most would agree that these are the respective gaps you expect between drivers pitted against each other on an average circuit in average conditions.
Really? How can I see that from your results? The only one who has said something about your calculated gaps is Ben24, and he actually felt that rb and fer were wrong. And so did I, and that was also the cause why I decided to fact check it.
Member
Posted 23 November 2022 - 15:11
First of all, there is no 'variation of mine' that I laid out here. I'm repeating myself when I say that what I did was verify the presented results by using your own method the way it was written down by you, the way I understood it, in particular the 3rd rule (as that's what's mainly causing the 92% difference in delta):
By elaborating further, it came to know that you didn't delete the 3 smallest and 3 biggest gaps, but that you deleted the 3 biggest gap-wins from each driver.
Secondly, no, using the 'delete the 3 biggest gaps' method (overall gaps, not 3 from each driver) I used to come to the 0.247s figure would certainly not mean disregarding 6 wins of driver A and keep the one of B. I'm actually amazed to read how you went from 3 to 6. Nowhere have I used '6', always talking about "3 biggest gaps", and in fact it is you who deletes 6 results, 3 of each driver. Anyway no, it would mean that B's win would be deleted, and 2 of A's. Like I said, outliers do not discriminate. So the result would be driver A wins with an avg of 0.1s per quali.
It turns out that in this instance, you get the same results as when you calculate it the way you used to come to this years' gaps. However, that's mainly because of choosing a convenient scenario, but it actually serves, to an extent, as an example of pointing at a flaw in your methodology which I implied earlier. Your method says: "Delete the 3 biggest wins of A, and the 3 biggest wins of B." However, driver B doesn't even have 3 wins, so why cement that as some kind of an axioma into a method? The way I calculated is superior because it doesn't discriminate with regard to outliers between drivers. If all the outliers happened to one driver, then so be it.
This brings me to the next point. If I were to want to paint me a representative picture of quali delta's, I would not set a limit of disregarding results at all, no minimum, no maximum. The fact that in the rb case my 0.247s figure is still representative by disregarding solely the 3 biggest gaps, is because the only two other quali's I'd disregard were quite evenly big and divided equally amongst the two dirvers. And again, if there were less than 3 anomalies, then I would delete less accordingly.
Ah OK, so your process involves subjectivity. That's fine. My suggestion is using a very basic statistical method that's the same for everybody, with the only chance of variation being what I deem a 'competive laptime in the same session' to be, but it shouldn't affect the results too much. Others seem to appreciate this method; fine if you don't. But when looking at trends in any sets of data, medians are an accepted means of doing it (no pun intended).
No, you're wrong. You just chose a convenient example, as I said above. You can get the exact same result by using a lot of different methods, including the method that was used by me to come to the 0.247s figure as I showed above.
I really think you might be misunderstanding my intent of this exercise.
These again, are erroneous statements. First off, max wasn't "82% of the time faster anyway", 4 of his 18 wins came precisely by those anomalies. Secondly, eliminating 75% of PER's wins is not "fine" at all, as 75% of his wins were on merit. I do not understand why somebody who says his aim is to eliminate fluke results and retain the 'legit' ones to come to a 'true picture', lets the fluke wins of one driver stand, while deleting the wins on merit of the other on top. A double whammy.Regarding the "a weekend where a driver wins doesn't matter in case of a trend", it does matter. A lot. Just look at this real life example in which your gap is 92% bigger than if you wouldn't have disregarded PER's wins on merit.
If one driver is faster than the other 8 times out of 10 with no anomalies, and the other driver gets a lucky day where they are faster, then 33% of their 'wins' in that set are due to luck. So whilst to you it looks disproportionate when that data is left out of the median calculation, it its just how it is.
Yeah, so I thought, But are you sure? Because you do seem to care a lot about this all, especially about whether the gap between the rb-pair is perceived to be big enough.
I'm not the one writing long posts disparaging someone else's calculations as 'wrong' in bold letters and not offering an alternative to show why. I'm not saying you're wrong (or anybody else is); just I'm presenting a different way of showing it that might be a more true reflection.
There's a difference between being surprised by something, and that thing being wrong. And when statistical methods are involved, there is rarely a wrong answer, unless mistakes are made in the source data.
I'm trying to explain that the whole point of my suggested method is to eliminate subjectivity (I really don't care about the gaps between drivers - I like Checo, so why would I want to put him down?). Yes sometimes someone will have really poor luck over more events which *might* have an effect, but I don't think its the case with the Red Bull example. Given you're the only one arguing against the numbers (so far) I think it does a pretty good job for most people.
Anyway, here's the file so you can have a look: https://easyupload.io/0n0t6c
(sorry its a bit messy as I wasn't intending to share, but the bold columns are what counts towards the average, with the red numbers left out (top 3 / bottom 3 conditionally formatted to indicate this)).
Edited by JimmyClark, 23 November 2022 - 15:21.
Member
Posted 23 November 2022 - 15:29
Thanks Tyker for theses stats. I enjoyed them
Cheers, it seems to have set up some rigorous debate as well.
Member
Posted 23 November 2022 - 15:37
Obviously this is a very simple example, but you can see from my results that most would agree that these are the respective gaps you expect between drivers pitted against each other on an average circuit in average conditions.
Oh, and about this
Really? How can I see that from your results? The only one who has said something about your calculated gaps is Ben24, and he actually felt that rb and fer were wrong. And so did I, and that was also the cause why I decided to fact check it.
I didn't want to get involved too much in this but I'm curious to what is meant by average conditions.
Edited by tyker, 23 November 2022 - 15:37.
Member
Posted 23 November 2022 - 15:49
I didn't want to get involved too much in this but I'm curious to what is meant by average conditions.
Hah sorry for hijacking; it wasn't intended I didn't realise my offer of an alternative to add to your excellent record keeping would be so controversial...
I just mean that if you are averaging out the times over a whole years worth of races, then indeed it will be 'average', but it could be a poor choice of wording there on my part. So essentially when I mean average conditions, I would say 'you would expect Teammate A to qualifying X.XXXs behind Teammate B on a very normal day on an average circuit'
(incidentally, this made me think what F1's most average circuit is - the mean qualifying session in 2022 with roughly 1m27 lap time, but this does include wet qualis too so the time is actually likely a bit less, and with the average circuit length of the calendar 5.151km, I think Abu Dhabi is the most typical circuit you can get with quali times around 1m24/25 and a length of 5.281km).
Edited by JimmyClark, 23 November 2022 - 15:50.
Member
Posted 23 November 2022 - 17:06
Hah sorry for hijacking; it wasn't intended
I just mean that if you are averaging out the times over a whole years worth of races, then indeed it will be 'average', but it could be a poor choice of wording there on my part. So essentially when I mean average conditions, I would say 'you would expect Teammate A to qualifying X.XXXs behind Teammate B on a very normal day on an average circuit'
(incidentally, this made me think what F1's most average circuit is - the mean qualifying session in 2022 with roughly 1m27 lap time, but this does include wet qualis too so the time is actually likely a bit less, and with the average circuit length of the calendar 5.151km, I think Abu Dhabi is the most typical circuit you can get with quali times around 1m24/25 and a length of 5.281km).
That's perfectly alright, people have come up with their own systems before which includes everything and then has mathematical solutions for outliers. while others have voiced opinion that everything should be included, but that's not my objective, things like wet weather driving is another skill again.
Thanks for explaining, so average conditions is a very normal day on an average circuit, that's what's confusing for me because I would see a very normal day for qualifying being one that is dry.
Member
Posted 23 November 2022 - 18:13
Ah OK, so your process involves subjectivity. That's fine. My suggestion is using a very basic statistical method that's the same for everybody, with the only chance of variation being what I deem a 'competive laptime in the same session' to be, but it shouldn't affect the results too much. Others seem to appreciate this method; fine if you don't. But when looking at trends in any sets of data, medians are an accepted means of doing it (no pun intended).
LOL, "subjectivity" coming from a fan who chooses to randomly delete the 3 best quali's of drivers. Why not 4, or 5? Could it be because that would mean a disadvantageous outcome for your favourite driver, as representative PER-wins would be disregarded? You're sooo transparent. Thát's subjectivity at its best.
Another thing, don't try to make this about me or my process in the first place. I didn't post a list of the delta's, even though it struck me that from the get go you tried to get such a thing from me, just for the sake of arguing anyone who doesn't adhere to the fake half a sec gap at rb I guess.
And about the perceived subjectivity, you're funny. There's nothing subjective at all in keeping the 3 PER's quali times intact; on the other hand, it is completely arbitrary and subjective to have them disregarded. Well subjective at least, because I'm convinced it wasn't arbitrary as you seem to have a narrative to tell.
Don't try to use backhand tactics like saying "Ah, OK blablabla" to insinuate I told you I was using subjective metrics. As everybody who can read, it is you utilizing illogical, subjective rules like "Hey, lets delete 75% of PERs wins" so the gap increases by 92%. I didn't omit any results you didn't omit yourself, and on top of that, you doubled them.
And cut the 'statistical crap' mantra, it won't work with me, and it's painful to watch. Just to let you know, I'm a STEM student, and if I were to use your believes about statistics that you were so kindly to share with us, and would go out in the open with it, I can guarantee you that my professors would be lobbying for the university to get me kicked out, and I'd be the laughing stock among my peers. Besides, if I were to use your 'statistics', I'd have never gotten into uni in the 1st place.
You seriously make no sense at all: "..but it shouldn't affect the results too much"
Really? I already showed that it causes the gap to grow with 92%, but keep on turning a blind eye, and repeating the same erroneous broken mantra by all means. Or are you arguing 92% isn't that much? Like I said, no sense at all.
Others appreciate your drivel? Who are those? And even if you were to find others who happen to lack the same basic knowledge as you do, what would that prove? That more people have a blind spot for analysis? That fan-ism supersedes reason? Please leave the "accepted means of doing"-mantra to those who are easily fooled. Your 'means of doing things' would have you get flunked in middle school already, let alone higher education.
I really think you might be misunderstanding my intent of this exercise.
I really really think I'm not. How could I? It says so in your OP:
I'll present my alternative. My methodology is:
- Take times from the last session both drivers set a flying lap in (to mitigate against changing conditions between Q1/Q2/Q3).
- The average laptime in 2022 was 87s, so all gaps are equalised to an 87s laptime, to try to make the gaps consistent when looking at, e.g., a dry Red Bull Ring vs a wet Singapore.
- The top and bottom 3 gaps between teammates are disregarded, to get a more median average. This should eliminate freak results / bad luck from the numbers. Thus the below is from the closest 16 races for each pairing.
The more I'm surprised to learn from you that you don't actually care for the bolded parts, because you per definition disregard the 3 top results from both drivers, regardless how they came to be. And if there were actually more than 3 "freak results", you'd not omit them. That's subjectivity right in front of you and you're deliberately choosing not to see it, and moreover, your chosen methodology doesn't fit your own stated intentions at all (while actually mine does, which is quite ironic). So it actually seems you forgot about 'your intentions' yourself.
Your assumption that 3 quali's for each driver "should" cover the freak results is simply wrong, even a 6 year old can see that, because what kind of law of nature dictates that every driver will have exactly 3 freak result wins, no more, no less? Tell that to the ALO-fans on this years' ALP-threads, or from '18 those of ALO again and RICs. I noticed you don't answer any of my (implicated) questions I asked you, and I'm still waiting for an answer why omitting Brazil was "controversial" according to you, while it literally made no difference at all as that was done by both. Again, it all makes no sense whatsoever.
If one driver is faster than the other 8 times out of 10 with no anomalies, and the other driver gets a lucky day where they are faster, then 33% of their 'wins' in that set are due to luck. So whilst to you it looks disproportionate when that data is left out of the median calculation, it its just how it is.
O this is getting better and better. Again, you're trying to attribute statements I never made, quite the opposite, to me.
And, how fitting, you made a calculation error. How can a driver that has lost 8 out of 10 times, have a win, or wins, that constitute 33% of his total wins? LOL, just stop already, it's getting really embarrassing.
I'm not the one writing long posts disparaging someone else's calculations as 'wrong' in bold letters and not offering an alternative to show why.
Yeah sure, Í'm the one disparaging calculations. Take a look again at how the conversation developed. And bold letters, really? The sole two, non-calculation BTW, instances were one to emphasize that I (eye) wouldn't set an arbitrary, subjective number of 'disregards' in advance, and another with your stated intentions, as it seemed to me you forgot about them (which has only been reinforced since your last post). And here you go again, trying to portray me in some kind of neg, this time not by putting words I never said in my mouth, but the other way around, claiming "not offering an alternative to show why". I gave you alternatives, lots of them, over and over (that's another reason for the bolded parts, everything I say or explain seems to fly right past by you), but if you choose to "disparage" or even completely ignore them, yeah, that's up to you.
And the long posts should offer you a clue about the reason why they are that long, its's because of explanations, alternatives and what not. Not really one should complain about.
I'm not saying you're wrong (or anybody else is); just I'm presenting a different way of showing it that might be a more true reflection.
You could've fooled me (well, you have then); yes, and I countered by that that might not be the case, but that didn't seem to be acceptable to you.
There's a difference between being surprised by something, and that thing being wrong. And when statistical methods are involved, there is rarely a wrong answer, unless mistakes are made in the source data..
What I really would like to happen, is for you to have a conversation with a stat prof, and for me to be a witness. How dare he flunk students, everything goes.
I'm trying to explain that the whole point of my suggested method is to eliminate subjectivity (I really don't care about the gaps between drivers - I like Checo, so why would I want to put him down?). Yes sometimes someone will have really poor luck over more events which *might* have an effect, but I don't think its the case with the Red Bull example. Given you're the only one arguing against the numbers (so far) I think it does a pretty good job for most people.
Like I said, your methodology, which is subjective to the bone, does not align with your alleged intentions. Why would you want to put him down? You tell me, but apart from the fact you almost doubled his gap to his teammate, you felt it to be necessary to very much highlight it and use wording with very much negative connotation:
Using this methodology, the teammate gaps were:
1. ALB > LAT = -0.497
2. VER > PER = -0.475
3. NOR > RIC = -0.384
4. MAG > SCH = -0.292
5. ALO > OCO = -0.264
6. BOT > ZHO = -0.250
7. GAS > TSU = -0.199
8. VET > STR = -0.152
9. HAM > RUS = -0.116
10. LEC > SAI = -0.046
The takeaways are...
- Perez is as abject in qualifying against Verstappen as Latifi is against Albon. It also means he is the outlier in that he is the only one of the top 4 worst performing qualifiers vs teammates to keep his drive for 2023.
- Leclerc is not the dominant qualifier over Sainz that many would assume, and were the closest two teammates.
- The Alpha Tauris and Aston Martins were pretty close, some might say surprisingly.
- Alonso and Hamilton aren't bad for old men.
So let me see: abject, comparison with LAT, one of the 4 'worst performers', and yet still keeping his drive, the only one of the 'worst-4'. I count around a quadruple dig. If thát's not putting someone down, nothing is. And if thát's how you talk about people you like, I'm curious about what you say about people you don't like.
And again you give reason to everyone with a brain to question whether you actually understand your own system. You talk about *might* have an effect. No, it will definitely have an effect, if one driver has more than your arbitrarily chosen cutoff of '3 best scores' freak results, unless the other driver has an equal 'freak result-delta' to counter with.
As for the rb this year not having that, untrue, but the impact isn't even that big. What ís big however, is you omitting results that weren't due to "poor luck", but perfectly normal results. So another big flaw in your used methodology to which you are simply unwilling to admit, even though I've pointed it about a dozen times now (see what I did there).
Well at least you're consistent. Again, you're using the 'nobody complains, so everybody agrees', classic false dichotomy. Nice. So if nobody disagrees with me too, everybody agrees with me disagreeing with you. This just goes on to reinforce my earlier assessment of your ability to reason.
Cheers.
And BTW, nice to know that you think Ben24 is "nobody".
Edited by eab, 23 November 2022 - 18:21.
Member
Posted 23 November 2022 - 19:31
eaB, I'm going to leave it there as its clear we are going round in circles, and you are now getting personal with me and it is derailing a sensible thread.
Edited by JimmyClark, 23 November 2022 - 19:34.
© MOTORSPORT NETWORK 2024. All rights reserved. Community Forum Software by IP.Board
