F1 2016 vs 2017: mathematical/statistical comparison

Post here all non technical related topics about Formula One. This includes race results, discussions, testing analysis etc. TV coverage and other personal questions should be in Off topic chat.
jonas_linder
3
Joined: 03 Mar 2016, 14:51

Re: F1 2016 vs 2017: mathematical/statistical comparison

Post

During the second test when the worst moaning was going on I was considering to just stop coming to this forum since it was more opinion than technical analysis! But then your preliminary analysis came out which lured me back and now with this complete analysis I'm glad that I did not go away!

Your analysis is very insightful and well thought through. I'm especially impressed about your ingenuity for compensating unknown factors! Furthermore, it is refreshing to see someone admitting that there is too much uncertainty to draw any conclusions.

=D> =D> =D> =D> =D> =D> =D> =D> =D> =D> =D> =D> =D> =D> =D>

Now to a question! There are certainly still a lot of unknown factors that are not considered, many of these can probably not be estimated by the limited datasets that are available. Are there other factors that you think can be compensated for using the available data?

Keep up the awesome work!

ripper
39
Joined: 26 Aug 2015, 22:19

Re: F1 2016 vs 2017: mathematical/statistical comparison

Post

12:42 1'21"127 per Vettel, che migliora il suo crono di riferimento ma rimane in quarta posizione - first lap medium stint
12:43 1'21"4 per Vettel nel secondo passaggio di questo run con gomme medium - second lap medium stint
12:44 1'21"6 per Vettel al terzo giro sulle medium - third lap medium stint

then it becomes less clear:
12:47 Si alzano i tempi di Vettel, che ora gira sul passo dell'1'22" basso -> Vettel laps slower : low 1'22
12:52 Vettel continua a martellare sull'1'22"3. Ora il tedesco fotocopia lo stesso tempo da giri -> Vettel is doing many laps with same time 1:22"3

User avatar
Mr.G
34
Joined: 10 Feb 2010, 22:52
Location: Slovakia

Re: F1 2016 vs 2017: mathematical/statistical comparison

Post

whatthefat - could you please try at leas some simplified comparison? I think from the times we have you should be able to create at least some rough data for the stint... Pretty please :D
Art without engineering is dreaming. Engineering without art is calculating. Steven K. Roberts

User avatar
Big Mangalhit
27
Joined: 03 Dec 2015, 15:39

Re: F1 2016 vs 2017: mathematical/statistical comparison

Post

Amazing analysis!!!! The best I've seen ever for pre season.

Just one question. I noticed that in the penultimate graph in which you plot different points according to the hypothetical initial fuel of the stint you have the same degradation tyre rate for any initial fuel level.

Maybe I am just confused but it seems the model assumes a linear degradation rate of tyres with any fuel level, but wouldn't it be expectable that a heavier loaded tyre would have a bigger degradation and thus to be corrected for that? Or is that value so small it doesn't matter? Or was that taken into account and I didn't see it?

User avatar
JonoNic
4
Joined: 05 Mar 2015, 15:54

Re: F1 2016 vs 2017: mathematical/statistical comparison

Post

Could we have it broken into sectors comparison too?

Sent from my SM-A700F using Tapatalk
Always find the gap then use it.

GoranF1
155
Joined: 16 Dec 2014, 12:53
Location: Zagreb,Croatia
Contact:

Re: F1 2016 vs 2017: mathematical/statistical comparison

Post

Dont know if this was posted before.....

https://f1metrics.wordpress.com
"I have no idols. I admire work, dedication & competence."

User avatar
whatthefat
0
Joined: 13 Sep 2014, 01:02
Contact:

Re: F1 2016 vs 2017: mathematical/statistical comparison

Post

ripper wrote:12:42 1'21"127 per Vettel, che migliora il suo crono di riferimento ma rimane in quarta posizione - first lap medium stint
12:43 1'21"4 per Vettel nel secondo passaggio di questo run con gomme medium - second lap medium stint
12:44 1'21"6 per Vettel al terzo giro sulle medium - third lap medium stint

then it becomes less clear:
12:47 Si alzano i tempi di Vettel, che ora gira sul passo dell'1'22" basso -> Vettel laps slower : low 1'22
12:52 Vettel continua a martellare sull'1'22"3. Ora il tedesco fotocopia lo stesso tempo da giri -> Vettel is doing many laps with same time 1:22"3
Aha, thank you! So then it does sound like some slight degradation occurring across the stint.

He started the stint with a 1:21.1 and 1:21.4 (reported at 12:42 and 12:43) then 1:21.6 (reported at 12:44), then went into low 1:22s (reported at 12:47) and apparently was consistently doing 1:22.3s (last 6 laps within a couple of tenths reported at 12:52). Don't know what happened after that, but it was later reported as a stint with "14 flying laps" at 13:12.

So, piecing that together, I think the stint looked something like this in the first 11 laps:
21.1, 21.4, 21.6, 21.9, 22.1, 22.2, 22.3, 22.2, 22.3, 22.4, 22.3, ...

Fitting those data, I get the following.

Base time = 81.3 sec
Fuel-corrected deg rate = 0.17 sec/lap

Compare this to the Mercedes stint at:

Base time = 81.3 sec
Fuel-corrected deg rate = 0.08 sec/lap

Again, we don't know the true fuel loads, so it's hard to make a direct comparison here, but the Mercedes stint remains very impressive.

User avatar
whatthefat
0
Joined: 13 Sep 2014, 01:02
Contact:

Re: F1 2016 vs 2017: mathematical/statistical comparison

Post

Big Mangalhit wrote:Amazing analysis!!!! The best I've seen ever for pre season.

Just one question. I noticed that in the penultimate graph in which you plot different points according to the hypothetical initial fuel of the stint you have the same degradation tyre rate for any initial fuel level.

Maybe I am just confused but it seems the model assumes a linear degradation rate of tyres with any fuel level, but wouldn't it be expectable that a heavier loaded tyre would have a bigger degradation and thus to be corrected for that? Or is that value so small it doesn't matter? Or was that taken into account and I didn't see it?
You are right, these fits are assuming the same deg rate at low and high fuel loads. In general, I would agree that heavier fuel loads ought to produce slightly higher deg rates, all else being equal, but I don't noticeably see that in the dataset.

User avatar
whatthefat
0
Joined: 13 Sep 2014, 01:02
Contact:

Re: F1 2016 vs 2017: mathematical/statistical comparison

Post

jonas_linder wrote:During the second test when the worst moaning was going on I was considering to just stop coming to this forum since it was more opinion than technical analysis! But then your preliminary analysis came out which lured me back and now with this complete analysis I'm glad that I did not go away!

Your analysis is very insightful and well thought through. I'm especially impressed about your ingenuity for compensating unknown factors! Furthermore, it is refreshing to see someone admitting that there is too much uncertainty to draw any conclusions.

=D> =D> =D> =D> =D> =D> =D> =D> =D> =D> =D> =D> =D> =D> =D>

Now to a question! There are certainly still a lot of unknown factors that are not considered, many of these can probably not be estimated by the limited datasets that are available. Are there other factors that you think can be compensated for using the available data?

Keep up the awesome work!
Thank you very much!

I think this is near the limit of being an unidentifiable model with the variables we're already including and the limited data we have available. People have suggested trying to correct for the day of the test or the air/track temperature, for instance. I'd love to do that, but I don't think it can be done from just these data. I would need some independent way of estimating these effects, because there are already too many other variables changing in concert.

You know, if we had GPS data as well, then we could do some very impressive things, because we could see cornering speeds, we could estimate engine modes from straight-line speeds, etc. That, I guess, is the domain of the teams, and they don't tend to share these data!

User avatar
dans79
267
Joined: 03 Mar 2013, 19:33
Location: USA

Re: F1 2016 vs 2017: mathematical/statistical comparison

Post

whatthefat wrote: You know, if we had GPS data as well, then we could do some very impressive things, because we could see cornering speeds, we could estimate engine modes from straight-line speeds, etc. That, I guess, is the domain of the teams, and they don't tend to share these data!
it would be an agonizingly manual process, but we could estimate some speeds and some accelerations by going frame by frame and looking at the distance traveled (width of the alternating curb segments,) over time (we know the frame rate of the video).
197 104 103 7

User avatar
Mr.G
34
Joined: 10 Feb 2010, 22:52
Location: Slovakia

Re: F1 2016 vs 2017: mathematical/statistical comparison

Post

whatthefat wrote:
ripper wrote:12:42 1'21"127 per Vettel, che migliora il suo crono di riferimento ma rimane in quarta posizione - first lap medium stint
12:43 1'21"4 per Vettel nel secondo passaggio di questo run con gomme medium - second lap medium stint
12:44 1'21"6 per Vettel al terzo giro sulle medium - third lap medium stint

then it becomes less clear:
12:47 Si alzano i tempi di Vettel, che ora gira sul passo dell'1'22" basso -> Vettel laps slower : low 1'22
12:52 Vettel continua a martellare sull'1'22"3. Ora il tedesco fotocopia lo stesso tempo da giri -> Vettel is doing many laps with same time 1:22"3
Aha, thank you! So then it does sound like some slight degradation occurring across the stint.

He started the stint with a 1:21.1 and 1:21.4 (reported at 12:42 and 12:43) then 1:21.6 (reported at 12:44), then went into low 1:22s (reported at 12:47) and apparently was consistently doing 1:22.3s (last 6 laps within a couple of tenths reported at 12:52). Don't know what happened after that, but it was later reported as a stint with "14 flying laps" at 13:12.

So, piecing that together, I think the stint looked something like this in the first 11 laps:
21.1, 21.4, 21.6, 21.9, 22.1, 22.2, 22.3, 22.2, 22.3, 22.4, 22.3, ...

Fitting those data, I get the following.

Base time = 81.3 sec
Fuel-corrected deg rate = 0.17 sec/lap

Compare this to the Mercedes stint at:

Base time = 81.3 sec
Fuel-corrected deg rate = 0.08 sec/lap

Again, we don't know the true fuel loads, so it's hard to make a direct comparison here, but the Mercedes stint remains very impressive.
Not the result I wanted to hear, but still thanks for the effort :)
Art without engineering is dreaming. Engineering without art is calculating. Steven K. Roberts

Restomaniac
0
Joined: 16 May 2016, 01:09
Location: Hull

Re: F1 2016 vs 2017: mathematical/statistical comparison

Post

Mr.G wrote:
whatthefat wrote:
ripper wrote:12:42 1'21"127 per Vettel, che migliora il suo crono di riferimento ma rimane in quarta posizione - first lap medium stint
12:43 1'21"4 per Vettel nel secondo passaggio di questo run con gomme medium - second lap medium stint
12:44 1'21"6 per Vettel al terzo giro sulle medium - third lap medium stint

then it becomes less clear:
12:47 Si alzano i tempi di Vettel, che ora gira sul passo dell'1'22" basso -> Vettel laps slower : low 1'22
12:52 Vettel continua a martellare sull'1'22"3. Ora il tedesco fotocopia lo stesso tempo da giri -> Vettel is doing many laps with same time 1:22"3
Aha, thank you! So then it does sound like some slight degradation occurring across the stint.

He started the stint with a 1:21.1 and 1:21.4 (reported at 12:42 and 12:43) then 1:21.6 (reported at 12:44), then went into low 1:22s (reported at 12:47) and apparently was consistently doing 1:22.3s (last 6 laps within a couple of tenths reported at 12:52). Don't know what happened after that, but it was later reported as a stint with "14 flying laps" at 13:12.

So, piecing that together, I think the stint looked something like this in the first 11 laps:
21.1, 21.4, 21.6, 21.9, 22.1, 22.2, 22.3, 22.2, 22.3, 22.4, 22.3, ...

Fitting those data, I get the following.

Base time = 81.3 sec
Fuel-corrected deg rate = 0.17 sec/lap

Compare this to the Mercedes stint at:

Base time = 81.3 sec
Fuel-corrected deg rate = 0.08 sec/lap

Again, we don't know the true fuel loads, so it's hard to make a direct comparison here, but the Mercedes stint remains very impressive.
Not the result I wanted to hear, but still thanks for the effort :)
Indeed it's even more depressing if as everyone expects Mercedes still have a better handle on qualifing engine modes.

User avatar
whatthefat
0
Joined: 13 Sep 2014, 01:02
Contact:

Re: F1 2016 vs 2017: mathematical/statistical comparison

Post

Mr.G wrote:
whatthefat wrote:
ripper wrote:12:42 1'21"127 per Vettel, che migliora il suo crono di riferimento ma rimane in quarta posizione - first lap medium stint
12:43 1'21"4 per Vettel nel secondo passaggio di questo run con gomme medium - second lap medium stint
12:44 1'21"6 per Vettel al terzo giro sulle medium - third lap medium stint

then it becomes less clear:
12:47 Si alzano i tempi di Vettel, che ora gira sul passo dell'1'22" basso -> Vettel laps slower : low 1'22
12:52 Vettel continua a martellare sull'1'22"3. Ora il tedesco fotocopia lo stesso tempo da giri -> Vettel is doing many laps with same time 1:22"3
Aha, thank you! So then it does sound like some slight degradation occurring across the stint.

He started the stint with a 1:21.1 and 1:21.4 (reported at 12:42 and 12:43) then 1:21.6 (reported at 12:44), then went into low 1:22s (reported at 12:47) and apparently was consistently doing 1:22.3s (last 6 laps within a couple of tenths reported at 12:52). Don't know what happened after that, but it was later reported as a stint with "14 flying laps" at 13:12.

So, piecing that together, I think the stint looked something like this in the first 11 laps:
21.1, 21.4, 21.6, 21.9, 22.1, 22.2, 22.3, 22.2, 22.3, 22.4, 22.3, ...

Fitting those data, I get the following.

Base time = 81.3 sec
Fuel-corrected deg rate = 0.17 sec/lap

Compare this to the Mercedes stint at:

Base time = 81.3 sec
Fuel-corrected deg rate = 0.08 sec/lap

Again, we don't know the true fuel loads, so it's hard to make a direct comparison here, but the Mercedes stint remains very impressive.
Not the result I wanted to hear, but still thanks for the effort :)
I'll just add a thought to this. If we assume the Mercedes was running at 7 laps of fuel for their 7-lap stint and the Ferrari was carrying 14 laps of fuel for their 14-lap stint, then these are the overlaid fuel-corrected times I get for the two stints, using my approximated times for Vettel from those reports.

Image

cokata
2
Joined: 16 May 2014, 19:50

Re: F1 2016 vs 2017: mathematical/statistical comparison

Post

Also one more thing to add for the long(isn) runs is that sometimes teams (for whatever reason) will use DRS, while others won't. That in itself can muddy the waters quite a bit.

ema00
1
Joined: 02 Mar 2012, 22:13

Re: F1 2016 vs 2017: mathematical/statistical comparison

Post

the worst thing is the lack of sim race in the second test by Mercedes, it would gave a better view on the situation