Page 5 of 9

Re: The Great Big 2021 A > B > C Thread

Posted: Sun Mar 28, 2021 2:52 am
by KingVoid
I have Verstappen vs Perez at 0.378%

90.659 divided by 90.318 equals 1.00378

But that’s just a minor issue.

Otherwise I do believe that most of these qualifying battles will fall within the predicted range, although I expect it to take time.

The only one that will likely be wrong is Vettel vs Stroll, because at this point I think Vettel is almost as finished as Raikkonen.

Re: The Great Big 2021 A > B > C Thread

Posted: Sun Mar 28, 2021 5:59 am
by Exediron
KingVoid wrote:
Sun Mar 28, 2021 2:52 am
I have Verstappen vs Perez at 0.378%

90.659 divided by 90.318 equals 1.00378
You're using a different equation than me, although I would have expected it to yield the same answer. I use the traditional long form of [V1-V2 / (V1+V2)/2] x 100, which produces a slightly different answer than your admittedly simpler solution.

90.318−90.659 = -0.341
90.318+90.659 / 2 = 90.4885
-0.341 / 90.4885 = -0.003768435

I don't round the resulting number, and just take the first three decimal places.

Re: The Great Big 2021 A > B > C Thread

Posted: Tue Mar 30, 2021 1:21 pm
by pokerman
Exediron wrote:
Sat Mar 27, 2021 9:21 pm
After one qualifying session, here is a very early look at how the lineups compare to the model predictions:

Leclerc / SAINZ
Projected: 0.386% to 0.542% (Leclerc)
Result (Q3): 0.595% (Leclerc)

VERSTAPPEN / PEREZ
Projected: 0.385% - 0.544% (Verstappen)
Result (Q2): 0.376% (Verstappen)

NORRIS / RICCIARDO
Projected: 0.262% - 0.403% (Ricciardo)
Result (Q3): 0.052% (Ricciardo)

STROLL / VETTEL
Projected: 0.262% - 0.316% (Vettel)
Result (Q1): 0.863% (Stroll)

OCON / ALONSO
Projected: 0.158% - 0.426% (Alonso)
Result (Q1): 0.938% (Alonso)

So after a single session, no single result falls within the predicted range. The Ferrari and Red Bull pairings are both very close to one end of the projection, while the McLaren pairing is much tighter than projected so far.

The Aston Martin and Alpine pairings threw up yellow flag affected results, so they shouldn't be taken to mean very much.
From the start you need to run the maths past me how does Leclerc being 0.331s quicker equate to 0.595% should it not be 0.369%?

Re: The Great Big 2021 A > B > C Thread

Posted: Tue Mar 30, 2021 1:33 pm
by A.J.
pokerman wrote:
Tue Mar 30, 2021 1:21 pm
Exediron wrote:
Sat Mar 27, 2021 9:21 pm
After one qualifying session, here is a very early look at how the lineups compare to the model predictions:

Leclerc / SAINZ
Projected: 0.386% to 0.542% (Leclerc)
Result (Q3): 0.595% (Leclerc)

VERSTAPPEN / PEREZ
Projected: 0.385% - 0.544% (Verstappen)
Result (Q2): 0.376% (Verstappen)

NORRIS / RICCIARDO
Projected: 0.262% - 0.403% (Ricciardo)
Result (Q3): 0.052% (Ricciardo)

STROLL / VETTEL
Projected: 0.262% - 0.316% (Vettel)
Result (Q1): 0.863% (Stroll)

OCON / ALONSO
Projected: 0.158% - 0.426% (Alonso)
Result (Q1): 0.938% (Alonso)

So after a single session, no single result falls within the predicted range. The Ferrari and Red Bull pairings are both very close to one end of the projection, while the McLaren pairing is much tighter than projected so far.

The Aston Martin and Alpine pairings threw up yellow flag affected results, so they shouldn't be taken to mean very much.
From the start you need to run the maths past me how does Leclerc being 0.331s quicker equate to 0.595% should it not be 0.369%?
Your math is wrong because your numbers are wrong. Leclerc did a 1:29.678, and Sainz did a 1:30.215 - so Leclerc was 0.537s ahead. No idea where you're getting your 0.331s number from.

Re: The Great Big 2021 A > B > C Thread

Posted: Tue Mar 30, 2021 1:39 pm
by JN23
pokerman wrote:
Tue Mar 30, 2021 1:21 pm
Exediron wrote:
Sat Mar 27, 2021 9:21 pm
After one qualifying session, here is a very early look at how the lineups compare to the model predictions:

Leclerc / SAINZ
Projected: 0.386% to 0.542% (Leclerc)
Result (Q3): 0.595% (Leclerc)

VERSTAPPEN / PEREZ
Projected: 0.385% - 0.544% (Verstappen)
Result (Q2): 0.376% (Verstappen)

NORRIS / RICCIARDO
Projected: 0.262% - 0.403% (Ricciardo)
Result (Q3): 0.052% (Ricciardo)

STROLL / VETTEL
Projected: 0.262% - 0.316% (Vettel)
Result (Q1): 0.863% (Stroll)

OCON / ALONSO
Projected: 0.158% - 0.426% (Alonso)
Result (Q1): 0.938% (Alonso)

So after a single session, no single result falls within the predicted range. The Ferrari and Red Bull pairings are both very close to one end of the projection, while the McLaren pairing is much tighter than projected so far.

The Aston Martin and Alpine pairings threw up yellow flag affected results, so they shouldn't be taken to mean very much.
From the start you need to run the maths past me how does Leclerc being 0.331s quicker equate to 0.595% should it not be 0.369%?
Lerclerc was 0.537s faster, not 0.331s.

Re: The Great Big 2021 A > B > C Thread

Posted: Tue Mar 30, 2021 1:52 pm
by A.J.
JN23 wrote:
Tue Mar 30, 2021 1:39 pm
pokerman wrote:
Tue Mar 30, 2021 1:21 pm
Exediron wrote:
Sat Mar 27, 2021 9:21 pm
After one qualifying session, here is a very early look at how the lineups compare to the model predictions:

Leclerc / SAINZ
Projected: 0.386% to 0.542% (Leclerc)
Result (Q3): 0.595% (Leclerc)

VERSTAPPEN / PEREZ
Projected: 0.385% - 0.544% (Verstappen)
Result (Q2): 0.376% (Verstappen)

NORRIS / RICCIARDO
Projected: 0.262% - 0.403% (Ricciardo)
Result (Q3): 0.052% (Ricciardo)

STROLL / VETTEL
Projected: 0.262% - 0.316% (Vettel)
Result (Q1): 0.863% (Stroll)

OCON / ALONSO
Projected: 0.158% - 0.426% (Alonso)
Result (Q1): 0.938% (Alonso)

So after a single session, no single result falls within the predicted range. The Ferrari and Red Bull pairings are both very close to one end of the projection, while the McLaren pairing is much tighter than projected so far.

The Aston Martin and Alpine pairings threw up yellow flag affected results, so they shouldn't be taken to mean very much.
From the start you need to run the maths past me how does Leclerc being 0.331s quicker equate to 0.595% should it not be 0.369%?
Lerclerc was 0.537s faster, not 0.331s.
I figured it out - he took Leclerc's Q3 time and compared it to Sainz's Q2 time (Sainz went slower in Q3) - but that's not how one does comparisons.

Re: The Great Big 2021 A > B > C Thread

Posted: Tue Mar 30, 2021 4:32 pm
by pokerman
A.J. wrote:
Tue Mar 30, 2021 1:33 pm
pokerman wrote:
Tue Mar 30, 2021 1:21 pm
Exediron wrote:
Sat Mar 27, 2021 9:21 pm
After one qualifying session, here is a very early look at how the lineups compare to the model predictions:

Leclerc / SAINZ
Projected: 0.386% to 0.542% (Leclerc)
Result (Q3): 0.595% (Leclerc)

VERSTAPPEN / PEREZ
Projected: 0.385% - 0.544% (Verstappen)
Result (Q2): 0.376% (Verstappen)

NORRIS / RICCIARDO
Projected: 0.262% - 0.403% (Ricciardo)
Result (Q3): 0.052% (Ricciardo)

STROLL / VETTEL
Projected: 0.262% - 0.316% (Vettel)
Result (Q1): 0.863% (Stroll)

OCON / ALONSO
Projected: 0.158% - 0.426% (Alonso)
Result (Q1): 0.938% (Alonso)

So after a single session, no single result falls within the predicted range. The Ferrari and Red Bull pairings are both very close to one end of the projection, while the McLaren pairing is much tighter than projected so far.

The Aston Martin and Alpine pairings threw up yellow flag affected results, so they shouldn't be taken to mean very much.
From the start you need to run the maths past me how does Leclerc being 0.331s quicker equate to 0.595% should it not be 0.369%?
Your math is wrong because your numbers are wrong. Leclerc did a 1:29.678, and Sainz did a 1:30.215 - so Leclerc was 0.537s ahead. No idea where you're getting your 0.331s number from.
Sainz went quicker in Q2.

Re: The Great Big 2021 A > B > C Thread

Posted: Tue Mar 30, 2021 4:36 pm
by pokerman
A.J. wrote:
Tue Mar 30, 2021 1:52 pm
JN23 wrote:
Tue Mar 30, 2021 1:39 pm
pokerman wrote:
Tue Mar 30, 2021 1:21 pm
Exediron wrote:
Sat Mar 27, 2021 9:21 pm
After one qualifying session, here is a very early look at how the lineups compare to the model predictions:

Leclerc / SAINZ
Projected: 0.386% to 0.542% (Leclerc)
Result (Q3): 0.595% (Leclerc)

VERSTAPPEN / PEREZ
Projected: 0.385% - 0.544% (Verstappen)
Result (Q2): 0.376% (Verstappen)

NORRIS / RICCIARDO
Projected: 0.262% - 0.403% (Ricciardo)
Result (Q3): 0.052% (Ricciardo)

STROLL / VETTEL
Projected: 0.262% - 0.316% (Vettel)
Result (Q1): 0.863% (Stroll)

OCON / ALONSO
Projected: 0.158% - 0.426% (Alonso)
Result (Q1): 0.938% (Alonso)

So after a single session, no single result falls within the predicted range. The Ferrari and Red Bull pairings are both very close to one end of the projection, while the McLaren pairing is much tighter than projected so far.

The Aston Martin and Alpine pairings threw up yellow flag affected results, so they shouldn't be taken to mean very much.
From the start you need to run the maths past me how does Leclerc being 0.331s quicker equate to 0.595% should it not be 0.369%?
Lerclerc was 0.537s faster, not 0.331s.
I figured it out - he took Leclerc's Q3 time and compared it to Sainz's Q2 time (Sainz went slower in Q3) - but that's not how one does comparisons.
Is it not we ignore what a driver is capable of for a slower lap, it might explain a few things.

Re: The Great Big 2021 A > B > C Thread

Posted: Tue Mar 30, 2021 4:37 pm
by A.J.
pokerman wrote:
Tue Mar 30, 2021 4:32 pm
A.J. wrote:
Tue Mar 30, 2021 1:33 pm
pokerman wrote:
Tue Mar 30, 2021 1:21 pm
Exediron wrote:
Sat Mar 27, 2021 9:21 pm
After one qualifying session, here is a very early look at how the lineups compare to the model predictions:

Leclerc / SAINZ
Projected: 0.386% to 0.542% (Leclerc)
Result (Q3): 0.595% (Leclerc)

VERSTAPPEN / PEREZ
Projected: 0.385% - 0.544% (Verstappen)
Result (Q2): 0.376% (Verstappen)

NORRIS / RICCIARDO
Projected: 0.262% - 0.403% (Ricciardo)
Result (Q3): 0.052% (Ricciardo)

STROLL / VETTEL
Projected: 0.262% - 0.316% (Vettel)
Result (Q1): 0.863% (Stroll)

OCON / ALONSO
Projected: 0.158% - 0.426% (Alonso)
Result (Q1): 0.938% (Alonso)

So after a single session, no single result falls within the predicted range. The Ferrari and Red Bull pairings are both very close to one end of the projection, while the McLaren pairing is much tighter than projected so far.

The Aston Martin and Alpine pairings threw up yellow flag affected results, so they shouldn't be taken to mean very much.
From the start you need to run the maths past me how does Leclerc being 0.331s quicker equate to 0.595% should it not be 0.369%?
Your math is wrong because your numbers are wrong. Leclerc did a 1:29.678, and Sainz did a 1:30.215 - so Leclerc was 0.537s ahead. No idea where you're getting your 0.331s number from.
Sainz went quicker in Q2.
That's irrelevant - Exideron's post that you quoted clearly states it is a Q3 comparison.

Re: The Great Big 2021 A > B > C Thread

Posted: Tue Mar 30, 2021 4:39 pm
by pokerman
A.J. wrote:
Tue Mar 30, 2021 4:37 pm
pokerman wrote:
Tue Mar 30, 2021 4:32 pm
A.J. wrote:
Tue Mar 30, 2021 1:33 pm
pokerman wrote:
Tue Mar 30, 2021 1:21 pm
Exediron wrote:
Sat Mar 27, 2021 9:21 pm
After one qualifying session, here is a very early look at how the lineups compare to the model predictions:

Leclerc / SAINZ
Projected: 0.386% to 0.542% (Leclerc)
Result (Q3): 0.595% (Leclerc)

VERSTAPPEN / PEREZ
Projected: 0.385% - 0.544% (Verstappen)
Result (Q2): 0.376% (Verstappen)

NORRIS / RICCIARDO
Projected: 0.262% - 0.403% (Ricciardo)
Result (Q3): 0.052% (Ricciardo)

STROLL / VETTEL
Projected: 0.262% - 0.316% (Vettel)
Result (Q1): 0.863% (Stroll)

OCON / ALONSO
Projected: 0.158% - 0.426% (Alonso)
Result (Q1): 0.938% (Alonso)

So after a single session, no single result falls within the predicted range. The Ferrari and Red Bull pairings are both very close to one end of the projection, while the McLaren pairing is much tighter than projected so far.

The Aston Martin and Alpine pairings threw up yellow flag affected results, so they shouldn't be taken to mean very much.
From the start you need to run the maths past me how does Leclerc being 0.331s quicker equate to 0.595% should it not be 0.369%?
Your math is wrong because your numbers are wrong. Leclerc did a 1:29.678, and Sainz did a 1:30.215 - so Leclerc was 0.537s ahead. No idea where you're getting your 0.331s number from.
Sainz went quicker in Q2.
That's irrelevant - Exideron's post that you quoted clearly states it is a Q3 comparison.
Then I just took it for granted that the method used in selecting laps was the same as mine.

Re: The Great Big 2021 A > B > C Thread

Posted: Tue Mar 30, 2021 4:43 pm
by A.J.
pokerman wrote:
Tue Mar 30, 2021 4:39 pm
A.J. wrote:
Tue Mar 30, 2021 4:37 pm
pokerman wrote:
Tue Mar 30, 2021 4:32 pm
A.J. wrote:
Tue Mar 30, 2021 1:33 pm
pokerman wrote:
Tue Mar 30, 2021 1:21 pm

From the start you need to run the maths past me how does Leclerc being 0.331s quicker equate to 0.595% should it not be 0.369%?
Your math is wrong because your numbers are wrong. Leclerc did a 1:29.678, and Sainz did a 1:30.215 - so Leclerc was 0.537s ahead. No idea where you're getting your 0.331s number from.
Sainz went quicker in Q2.
That's irrelevant - Exideron's post that you quoted clearly states it is a Q3 comparison.
Then I just took it for granted that the method used in selecting laps was the same as mine.
But why would you? The lap that decides your position is the lap in Q3 - the only reason to take a Q1 or Q2 lap is if the comparison for Q3 cannot be made for whatever reason. Q3 is supposed to be their ultimate pace - if a driver bottles it, then it is entirely on the driver.

Re: The Great Big 2021 A > B > C Thread

Posted: Tue Mar 30, 2021 4:45 pm
by pokerman
A.J. wrote:
Tue Mar 30, 2021 4:43 pm
pokerman wrote:
Tue Mar 30, 2021 4:39 pm
A.J. wrote:
Tue Mar 30, 2021 4:37 pm
pokerman wrote:
Tue Mar 30, 2021 4:32 pm
A.J. wrote:
Tue Mar 30, 2021 1:33 pm


Your math is wrong because your numbers are wrong. Leclerc did a 1:29.678, and Sainz did a 1:30.215 - so Leclerc was 0.537s ahead. No idea where you're getting your 0.331s number from.
Sainz went quicker in Q2.
That's irrelevant - Exideron's post that you quoted clearly states it is a Q3 comparison.
Then I just took it for granted that the method used in selecting laps was the same as mine.
But why would you? The lap that decides your position is the lap in Q3 - the only reason to take a Q1 or Q2 lap is if the comparison for Q3 cannot be made for whatever reason. Q3 is supposed to be their ultimate pace - if a driver bottles it, then it is entirely on the driver.
How can it be taken as their ultimate pace when they already showed that they could go quicker?

Re: The Great Big 2021 A > B > C Thread

Posted: Tue Mar 30, 2021 4:54 pm
by A.J.
pokerman wrote:
Tue Mar 30, 2021 4:45 pm
A.J. wrote:
Tue Mar 30, 2021 4:43 pm
pokerman wrote:
Tue Mar 30, 2021 4:39 pm
A.J. wrote:
Tue Mar 30, 2021 4:37 pm
pokerman wrote:
Tue Mar 30, 2021 4:32 pm

Sainz went quicker in Q2.
That's irrelevant - Exideron's post that you quoted clearly states it is a Q3 comparison.
Then I just took it for granted that the method used in selecting laps was the same as mine.
But why would you? The lap that decides your position is the lap in Q3 - the only reason to take a Q1 or Q2 lap is if the comparison for Q3 cannot be made for whatever reason. Q3 is supposed to be their ultimate pace - if a driver bottles it, then it is entirely on the driver.
How can it be taken as their ultimate pace when they already showed that they could go quicker?
Q3 is supposed to be their ultimate pace. If a driver buckles under pressure and goes slower, then of course that should count against them. The only reason to not consider Q3 times would be if the driver is unable to participate or has some issue preventing them from setting a representative lap time. Bottling it is not a reason.

Imagine if a driver goes faster in FP3 than in quali - would you consider that time when making a comparison with their teammate? Of course not, as that would be ridiculous.

Re: The Great Big 2021 A > B > C Thread

Posted: Tue Mar 30, 2021 6:05 pm
by pokerman
A.J. wrote:
Tue Mar 30, 2021 4:54 pm
pokerman wrote:
Tue Mar 30, 2021 4:45 pm
A.J. wrote:
Tue Mar 30, 2021 4:43 pm
pokerman wrote:
Tue Mar 30, 2021 4:39 pm
A.J. wrote:
Tue Mar 30, 2021 4:37 pm


That's irrelevant - Exideron's post that you quoted clearly states it is a Q3 comparison.
Then I just took it for granted that the method used in selecting laps was the same as mine.
But why would you? The lap that decides your position is the lap in Q3 - the only reason to take a Q1 or Q2 lap is if the comparison for Q3 cannot be made for whatever reason. Q3 is supposed to be their ultimate pace - if a driver bottles it, then it is entirely on the driver.
How can it be taken as their ultimate pace when they already showed that they could go quicker?
Q3 is supposed to be their ultimate pace. If a driver buckles under pressure and goes slower, then of course that should count against them. The only reason to not consider Q3 times would be if the driver is unable to participate or has some issue preventing them from setting a representative lap time. Bottling it is not a reason.

Imagine if a driver goes faster in FP3 than in quali - would you consider that time when making a comparison with their teammate? Of course not, as that would be ridiculous.
Well for a start it's qualifying stats and not practice stats, I've managed to find their 4 soft tyre runs:-

Leclerc
1:30.691
1:31.052
1:30.014
1:29.678

Sainz
1:30.861
1:30.009
1:30.536
1:30.215

So we are comparing Leclerc's fastest time with Sainz's 2nd fastest time, when trying to determine the ultimate pace of drivers why are we restricting some of the data which I think is relevant it's certainly can't be considered as outliers, is this how scientific analysis is normally conducted?

I didn't realise we were measuring who cracks the most under pressure as opposed to the more data collected the more accurate the measurement of ultimate speed.

Re: The Great Big 2021 A > B > C Thread

Posted: Tue Mar 30, 2021 7:10 pm
by JN23
Poker - what do you do if one driver gets knocked out in Q2 and the other Q3? Do you use both Q2 times or the Q2 time for the one who got knocked out and Q3 time for the other?

Re: The Great Big 2021 A > B > C Thread

Posted: Tue Mar 30, 2021 8:12 pm
by mikeyg123
A.J. wrote:
Tue Mar 30, 2021 4:54 pm
pokerman wrote:
Tue Mar 30, 2021 4:45 pm
A.J. wrote:
Tue Mar 30, 2021 4:43 pm
pokerman wrote:
Tue Mar 30, 2021 4:39 pm
A.J. wrote:
Tue Mar 30, 2021 4:37 pm


That's irrelevant - Exideron's post that you quoted clearly states it is a Q3 comparison.
Then I just took it for granted that the method used in selecting laps was the same as mine.
But why would you? The lap that decides your position is the lap in Q3 - the only reason to take a Q1 or Q2 lap is if the comparison for Q3 cannot be made for whatever reason. Q3 is supposed to be their ultimate pace - if a driver bottles it, then it is entirely on the driver.
How can it be taken as their ultimate pace when they already showed that they could go quicker?
Q3 is supposed to be their ultimate pace. If a driver buckles under pressure and goes slower, then of course that should count against them. The only reason to not consider Q3 times would be if the driver is unable to participate or has some issue preventing them from setting a representative lap time. Bottling it is not a reason.

Imagine if a driver goes faster in FP3 than in quali - would you consider that time when making a comparison with their teammate? Of course not, as that would be ridiculous.
I think your misunderstanding what these stats are supposed to represent. It's a drivers ultimate maximum pace in qualifying. Nothing more or less. It isn't meant to account for their ability to not crack under pressure or anything else.

With that in mind treat them as you see fit. Personally I find the work of Poker and other on this thread very useful and illuminating.

Re: The Great Big 2021 A > B > C Thread

Posted: Tue Mar 30, 2021 8:58 pm
by Exediron
It's a valid point concerning Sainz going slower on his Q3 lap, and I considered adding a note that he posted a quicker lap in Q2. My methodology, however, is to only use same session times. There is not always a linear improvement between sessions, so simply taking the fastest time across the whole of qualifying invites occasions where the track actually got slower.

I think it's entirely valid to take a faster lap from Q2 if there's no reason to believe the track got slower in Q3, but it isn't what I do.

Re: The Great Big 2021 A > B > C Thread

Posted: Wed Mar 31, 2021 9:29 am
by A.J.
pokerman wrote:
Tue Mar 30, 2021 6:05 pm
A.J. wrote:
Tue Mar 30, 2021 4:54 pm
pokerman wrote:
Tue Mar 30, 2021 4:45 pm
A.J. wrote:
Tue Mar 30, 2021 4:43 pm
pokerman wrote:
Tue Mar 30, 2021 4:39 pm

Then I just took it for granted that the method used in selecting laps was the same as mine.
But why would you? The lap that decides your position is the lap in Q3 - the only reason to take a Q1 or Q2 lap is if the comparison for Q3 cannot be made for whatever reason. Q3 is supposed to be their ultimate pace - if a driver bottles it, then it is entirely on the driver.
How can it be taken as their ultimate pace when they already showed that they could go quicker?
Q3 is supposed to be their ultimate pace. If a driver buckles under pressure and goes slower, then of course that should count against them. The only reason to not consider Q3 times would be if the driver is unable to participate or has some issue preventing them from setting a representative lap time. Bottling it is not a reason.

Imagine if a driver goes faster in FP3 than in quali - would you consider that time when making a comparison with their teammate? Of course not, as that would be ridiculous.
Well for a start it's qualifying stats and not practice stats, I've managed to find their 4 soft tyre runs:-

Leclerc
1:30.691
1:31.052
1:30.014
1:29.678

Sainz
1:30.861
1:30.009
1:30.536
1:30.215

So we are comparing Leclerc's fastest time with Sainz's 2nd fastest time, when trying to determine the ultimate pace of drivers why are we restricting some of the data which I think is relevant it's certainly can't be considered as outliers, is this how scientific analysis is normally conducted?

I didn't realise we were measuring who cracks the most under pressure as opposed to the more data collected the more accurate the measurement of ultimate speed.
Either it is the driver's fastest pace or it is not - if you're comparing Q2 to Q3, someone else can compare FP3 to Q3 - that just doesn't make sense. An apples-to-apples comparison is only possible when it is in the exact same session - unless it is not possible due to reasons I described in my previous post.

P.S. I didn't think this was "scientific analysis", but if you were doing that you would realize that using inconsistent data (with 2 different sources) renders the comparative study invalid. Heck, in testing Sainz did 1:29.611 on softs (faster than Leclerc in Q3), so now let's consider that as his "ultimate pace"?

Re: The Great Big 2021 A > B > C Thread

Posted: Wed Mar 31, 2021 9:36 am
by A.J.
mikeyg123 wrote:
Tue Mar 30, 2021 8:12 pm
A.J. wrote:
Tue Mar 30, 2021 4:54 pm
pokerman wrote:
Tue Mar 30, 2021 4:45 pm
A.J. wrote:
Tue Mar 30, 2021 4:43 pm
pokerman wrote:
Tue Mar 30, 2021 4:39 pm

Then I just took it for granted that the method used in selecting laps was the same as mine.
But why would you? The lap that decides your position is the lap in Q3 - the only reason to take a Q1 or Q2 lap is if the comparison for Q3 cannot be made for whatever reason. Q3 is supposed to be their ultimate pace - if a driver bottles it, then it is entirely on the driver.
How can it be taken as their ultimate pace when they already showed that they could go quicker?
Q3 is supposed to be their ultimate pace. If a driver buckles under pressure and goes slower, then of course that should count against them. The only reason to not consider Q3 times would be if the driver is unable to participate or has some issue preventing them from setting a representative lap time. Bottling it is not a reason.

Imagine if a driver goes faster in FP3 than in quali - would you consider that time when making a comparison with their teammate? Of course not, as that would be ridiculous.
I think your misunderstanding what these stats are supposed to represent. It's a drivers ultimate maximum pace in qualifying. Nothing more or less. It isn't meant to account for their ability to not crack under pressure or anything else.

With that in mind treat them as you see fit. Personally I find the work of Poker and other on this thread very useful and illuminating.
I understand that - but I think it is not a fair comparison if we don't use Q3 times when they are available and valid. I am not suggesting we are trying to measure their ability under pressure - but if someone underperforms due to lack of that ability, then it is what it is. For valid comparisons we should use the fastest time across the same session, instead of fastest time during the weekend.

I'm in agreement with you on the last bit :thumbup:

Re: The Great Big 2021 A > B > C Thread

Posted: Wed Mar 31, 2021 12:09 pm
by pokerman
JN23 wrote:
Tue Mar 30, 2021 7:10 pm
Poker - what do you do if one driver gets knocked out in Q2 and the other Q3? Do you use both Q2 times or the Q2 time for the one who got knocked out and Q3 time for the other?
I then just use the Q1 and Q2 times nominally Q3 times can be quicker as the grip ramps up also of course the driver in Q3 has had more opportunity to improve his time.

Re: The Great Big 2021 A > B > C Thread

Posted: Wed Mar 31, 2021 12:10 pm
by pokerman
mikeyg123 wrote:
Tue Mar 30, 2021 8:12 pm
A.J. wrote:
Tue Mar 30, 2021 4:54 pm
pokerman wrote:
Tue Mar 30, 2021 4:45 pm
A.J. wrote:
Tue Mar 30, 2021 4:43 pm
pokerman wrote:
Tue Mar 30, 2021 4:39 pm

Then I just took it for granted that the method used in selecting laps was the same as mine.
But why would you? The lap that decides your position is the lap in Q3 - the only reason to take a Q1 or Q2 lap is if the comparison for Q3 cannot be made for whatever reason. Q3 is supposed to be their ultimate pace - if a driver bottles it, then it is entirely on the driver.
How can it be taken as their ultimate pace when they already showed that they could go quicker?
Q3 is supposed to be their ultimate pace. If a driver buckles under pressure and goes slower, then of course that should count against them. The only reason to not consider Q3 times would be if the driver is unable to participate or has some issue preventing them from setting a representative lap time. Bottling it is not a reason.

Imagine if a driver goes faster in FP3 than in quali - would you consider that time when making a comparison with their teammate? Of course not, as that would be ridiculous.
I think your misunderstanding what these stats are supposed to represent. It's a drivers ultimate maximum pace in qualifying. Nothing more or less. It isn't meant to account for their ability to not crack under pressure or anything else.

With that in mind treat them as you see fit. Personally I find the work of Poker and other on this thread very useful and illuminating.
Thank you. :thumbup:

Re: The Great Big 2021 A > B > C Thread

Posted: Wed Mar 31, 2021 12:16 pm
by pokerman
A.J. wrote:
Wed Mar 31, 2021 9:29 am
pokerman wrote:
Tue Mar 30, 2021 6:05 pm
A.J. wrote:
Tue Mar 30, 2021 4:54 pm
pokerman wrote:
Tue Mar 30, 2021 4:45 pm
A.J. wrote:
Tue Mar 30, 2021 4:43 pm


But why would you? The lap that decides your position is the lap in Q3 - the only reason to take a Q1 or Q2 lap is if the comparison for Q3 cannot be made for whatever reason. Q3 is supposed to be their ultimate pace - if a driver bottles it, then it is entirely on the driver.
How can it be taken as their ultimate pace when they already showed that they could go quicker?
Q3 is supposed to be their ultimate pace. If a driver buckles under pressure and goes slower, then of course that should count against them. The only reason to not consider Q3 times would be if the driver is unable to participate or has some issue preventing them from setting a representative lap time. Bottling it is not a reason.

Imagine if a driver goes faster in FP3 than in quali - would you consider that time when making a comparison with their teammate? Of course not, as that would be ridiculous.
Well for a start it's qualifying stats and not practice stats, I've managed to find their 4 soft tyre runs:-

Leclerc
1:30.691
1:31.052
1:30.014
1:29.678

Sainz
1:30.861
1:30.009
1:30.536
1:30.215

So we are comparing Leclerc's fastest time with Sainz's 2nd fastest time, when trying to determine the ultimate pace of drivers why are we restricting some of the data which I think is relevant it's certainly can't be considered as outliers, is this how scientific analysis is normally conducted?

I didn't realise we were measuring who cracks the most under pressure as opposed to the more data collected the more accurate the measurement of ultimate speed.
Either it is the driver's fastest pace or it is not - if you're comparing Q2 to Q3, someone else can compare FP3 to Q3 - that just doesn't make sense. An apples-to-apples comparison is only possible when it is in the exact same session - unless it is not possible due to reasons I described in my previous post.

P.S. I didn't think this was "scientific analysis", but if you were doing that you would realize that using inconsistent data (with 2 different sources) renders the comparative study invalid. Heck, in testing Sainz did 1:29.611 on softs (faster than Leclerc in Q3), so now let's consider that as his "ultimate pace"?
I'm not sure that drivers are looking to set the lap of their lives in either testing or practice when it counts for nothing, also I think you can appreciate that track conditions can change.

Re: The Great Big 2021 A > B > C Thread

Posted: Wed Mar 31, 2021 7:45 pm
by Paolo_Lasardi
A.J. wrote:
Wed Mar 31, 2021 9:29 am
pokerman wrote:
Tue Mar 30, 2021 6:05 pm
A.J. wrote:
Tue Mar 30, 2021 4:54 pm
pokerman wrote:
Tue Mar 30, 2021 4:45 pm
A.J. wrote:
Tue Mar 30, 2021 4:43 pm


But why would you? The lap that decides your position is the lap in Q3 - the only reason to take a Q1 or Q2 lap is if the comparison for Q3 cannot be made for whatever reason. Q3 is supposed to be their ultimate pace - if a driver bottles it, then it is entirely on the driver.
How can it be taken as their ultimate pace when they already showed that they could go quicker?
Q3 is supposed to be their ultimate pace. If a driver buckles under pressure and goes slower, then of course that should count against them. The only reason to not consider Q3 times would be if the driver is unable to participate or has some issue preventing them from setting a representative lap time. Bottling it is not a reason.

Imagine if a driver goes faster in FP3 than in quali - would you consider that time when making a comparison with their teammate? Of course not, as that would be ridiculous.
Well for a start it's qualifying stats and not practice stats, I've managed to find their 4 soft tyre runs:-

Leclerc
1:30.691
1:31.052
1:30.014
1:29.678

Sainz
1:30.861
1:30.009
1:30.536
1:30.215

So we are comparing Leclerc's fastest time with Sainz's 2nd fastest time, when trying to determine the ultimate pace of drivers why are we restricting some of the data which I think is relevant it's certainly can't be considered as outliers, is this how scientific analysis is normally conducted?

I didn't realise we were measuring who cracks the most under pressure as opposed to the more data collected the more accurate the measurement of ultimate speed.
Either it is the driver's fastest pace or it is not - if you're comparing Q2 to Q3, someone else can compare FP3 to Q3 - that just doesn't make sense. An apples-to-apples comparison is only possible when it is in the exact same session - unless it is not possible due to reasons I described in my previous post.

P.S. I didn't think this was "scientific analysis", but if you were doing that you would realize that using inconsistent data (with 2 different sources) renders the comparative study invalid. Heck, in testing Sainz did 1:29.611 on softs (faster than Leclerc in Q3), so now let's consider that as his "ultimate pace"?
I disagree. A scientific approach will try to use the data, which proxies the most similar circumstances. While this preferably will be from the same session, the case in question represents an exception: if a driver botches up his Q3 lap but had a clean Q2 lap, then the Q2 lap is more similar to the counterfactual data that is aimed for, i.e. a clean Q3 lap, than the botched Q3 lap. Using the botched Q3 lap produces a bigger mistake here than using the clean Q2 lap, from a scientific point of view.

However, the (inevitable) mistake could be further reduced if the clean Q2 lap in our case is corrected by the average change between clean Q2 and Q3 laps ( of the other drivers) in this event. Yet, this would require that the distribution of the changes is not too high, i.e. the average is meaningful.

Re: The Great Big 2021 A > B > C Thread

Posted: Thu Apr 01, 2021 11:34 am
by A.J.
Paolo_Lasardi wrote:
Wed Mar 31, 2021 7:45 pm
A.J. wrote:
Wed Mar 31, 2021 9:29 am
pokerman wrote:
Tue Mar 30, 2021 6:05 pm
A.J. wrote:
Tue Mar 30, 2021 4:54 pm
pokerman wrote:
Tue Mar 30, 2021 4:45 pm

How can it be taken as their ultimate pace when they already showed that they could go quicker?
Q3 is supposed to be their ultimate pace. If a driver buckles under pressure and goes slower, then of course that should count against them. The only reason to not consider Q3 times would be if the driver is unable to participate or has some issue preventing them from setting a representative lap time. Bottling it is not a reason.

Imagine if a driver goes faster in FP3 than in quali - would you consider that time when making a comparison with their teammate? Of course not, as that would be ridiculous.
Well for a start it's qualifying stats and not practice stats, I've managed to find their 4 soft tyre runs:-

Leclerc
1:30.691
1:31.052
1:30.014
1:29.678

Sainz
1:30.861
1:30.009
1:30.536
1:30.215

So we are comparing Leclerc's fastest time with Sainz's 2nd fastest time, when trying to determine the ultimate pace of drivers why are we restricting some of the data which I think is relevant it's certainly can't be considered as outliers, is this how scientific analysis is normally conducted?

I didn't realise we were measuring who cracks the most under pressure as opposed to the more data collected the more accurate the measurement of ultimate speed.
Either it is the driver's fastest pace or it is not - if you're comparing Q2 to Q3, someone else can compare FP3 to Q3 - that just doesn't make sense. An apples-to-apples comparison is only possible when it is in the exact same session - unless it is not possible due to reasons I described in my previous post.

P.S. I didn't think this was "scientific analysis", but if you were doing that you would realize that using inconsistent data (with 2 different sources) renders the comparative study invalid. Heck, in testing Sainz did 1:29.611 on softs (faster than Leclerc in Q3), so now let's consider that as his "ultimate pace"?
I disagree. A scientific approach will try to use the data, which proxies the most similar circumstances. While this preferably will be from the same session, the case in question represents an exception: if a driver botches up his Q3 lap but had a clean Q2 lap, then the Q2 lap is more similar to the counterfactual data that is aimed for, i.e. a clean Q3 lap, than the botched Q3 lap. Using the botched Q3 lap produces a bigger mistake here than using the clean Q2 lap, from a scientific point of view.

However, the (inevitable) mistake could be further reduced if the clean Q2 lap in our case is corrected by the average change between clean Q2 and Q3 laps ( of the other drivers) in this event. Yet, this would require that the distribution of the changes is not too high, i.e. the average is meaningful.
Fair enough :thumbup: - although please note that I'm not saying a "botched" lap should count if it's in the same session. If the delta is within reason it isn't a botched lap, it's just the driver unable to replicate his fastest time. I think counting times across sessions isn't a fair comparison, but I can see some people view it differently - all good.

Re: The Great Big 2021 A > B > C Thread

Posted: Thu Apr 01, 2021 12:27 pm
by pokerman
Exediron wrote:
Tue Mar 30, 2021 8:58 pm
It's a valid point concerning Sainz going slower on his Q3 lap, and I considered adding a note that he posted a quicker lap in Q2. My methodology, however, is to only use same session times. There is not always a linear improvement between sessions, so simply taking the fastest time across the whole of qualifying invites occasions where the track actually got slower.

I think it's entirely valid to take a faster lap from Q2 if there's no reason to believe the track got slower in Q3, but it isn't what I do.
Fair enough you have your own methodology whereas I prefer to be more hands on, at the end of the day we don't seem to be that far apart most of the time.

It will be interesting to see what we come up with after this year, for my part for the first time I'm going to use percentages as I have to recognise that it's slightly more accurate, but then convert back to seconds because I've no concept of a percentage, a driver being 0.2s quicker sits better for me than saying a driver is 0.2% quicker which doesn't convert back to 0.2s but something less than that.

Re: The Great Big 2021 A > B > C Thread

Posted: Sun Apr 04, 2021 4:54 pm
by Johnson
Comparing Sainz Q2 vs Leclerc Q3 is the best example of both drivers at there best. For me, the most useful information.

Comparing Sainz Q3 vs Leclerc Q3 is an example of a mistake ridden lap of Sainz vs Leclerc's maximum.

Unless Sainz makes errors continually then surely the first comparison is not only more reflective of ability but also more useful to project future results. The conditions in this Q session were also consistent.

This raises the question, if we were racing at Bahrain again this weekend, which comparison do you think would be closer to the Q3 result? Again, I would go with comparison one.

"best sectors" is usually the most accurate indicator of outright pace, I think this had Leclerc 0.300 ahead.

Re: The Great Big 2021 A > B > C Thread

Posted: Wed Apr 07, 2021 3:15 am
by Tufty
Johnson wrote:
Sun Apr 04, 2021 4:54 pm
"best sectors" is usually the most accurate indicator of outright pace, I think this had Leclerc 0.300 ahead.
No. Because with electrical deployment used on different laps at different points by different drivers, you're not always comparing like for like.

Re: The Great Big 2021 A > B > C Thread

Posted: Wed Apr 07, 2021 11:32 am
by Johnson
Tufty wrote:
Wed Apr 07, 2021 3:15 am
Johnson wrote:
Sun Apr 04, 2021 4:54 pm
"best sectors" is usually the most accurate indicator of outright pace, I think this had Leclerc 0.300 ahead.
No. Because with electrical deployment used on different laps at different points by different drivers, you're not always comparing like for like.
Electrical deployment in qualifying is pre-programmed for the best lap time. It can be over ridden but it would make no sense to be deploying at different points - there is only one way of deployment that gives the best lap time and they have discovered that long before Q3.

Re: The Great Big 2021 A > B > C Thread

Posted: Mon Apr 19, 2021 5:52 am
by Exediron
Update from Imola:

Leclerc / SAINZ
Projected: 0.386% to 0.542% (Leclerc)
Current Result: 0.557% (Leclerc)
Data (click to show)

VERSTAPPEN / PEREZ
Projected: 0.385% - 0.544% (Verstappen)
Current Result: 0.153% (Verstappen)
Data (click to show)

NORRIS / RICCIARDO
Projected: 0.262% - 0.403% (Ricciardo)
Current Result: 0.058% (Ricciardo)
Data (click to show)

STROLL / VETTEL
Projected: 0.262% - 0.316% (Vettel)
Current Result: 0.601% (Stroll)
Data (click to show)

OCON / ALONSO
Projected: 0.158% - 0.426% (Alonso)
Current Result: 0.154% (Alonso)
Data (click to show)

Keep in mind that this is only two data points each, so the data is still highly subject to change. At present, I'm computing using the average of the two results; after the next race, I will switch to the median method, which will likely cause large initial changes from this race to the next. The result will become much more accurate the more data gets fed into it.

Currently, the Sainz/Leclerc and Ocon/Alonso head-to-heads are trending at something close to the indicated range, while the Perez/Verstappen and Norris/Ricciardo pairings are in the right direction but off the projected number. Only the Stroll/Vettel match-up is actually trending in the opposite direction to predictions.

Personally, I would say that the biggest inaccuracy is the McLaren pair, which indicates a slight edge for Ricciardo at the moment. The reality has been a strong edge for Norris, particularly in race pace. If Ricciardo gets on top of Norris over the course of the season as expected, his lucky early results may end up weighing the eventual result artificially in his direction.

Re: The Great Big 2021 A > B > C Thread

Posted: Mon Apr 19, 2021 5:18 pm
by pokerman
Exediron wrote:
Mon Apr 19, 2021 5:52 am
Update from Imola:

Leclerc / SAINZ
Projected: 0.386% to 0.542% (Leclerc)
Current Result: 0.557% (Leclerc)
Data (click to show)

VERSTAPPEN / PEREZ
Projected: 0.385% - 0.544% (Verstappen)
Current Result: 0.153% (Verstappen)
Data (click to show)

NORRIS / RICCIARDO
Projected: 0.262% - 0.403% (Ricciardo)
Current Result: 0.058% (Ricciardo)
Data (click to show)

STROLL / VETTEL
Projected: 0.262% - 0.316% (Vettel)
Current Result: 0.601% (Stroll)
Data (click to show)

OCON / ALONSO
Projected: 0.158% - 0.426% (Alonso)
Current Result: 0.154% (Alonso)
Data (click to show)

Keep in mind that this is only two data points each, so the data is still highly subject to change. At present, I'm computing using the average of the two results; after the next race, I will switch to the median method, which will likely cause large initial changes from this race to the next. The result will become much more accurate the more data gets fed into it.

Currently, the Sainz/Leclerc and Ocon/Alonso head-to-heads are trending at something close to the indicated range, while the Perez/Verstappen and Norris/Ricciardo pairings are in the right direction but off the projected number. Only the Stroll/Vettel match-up is actually trending in the opposite direction to predictions.

Personally, I would say that the biggest inaccuracy is the McLaren pair, which indicates a slight edge for Ricciardo at the moment. The reality has been a strong edge for Norris, particularly in race pace. If Ricciardo gets on top of Norris over the course of the season as expected, his lucky early results may end up weighing the eventual result artificially in his direction.
I think you're going to need the median to over write some of the Q1 comparisons from Bahrain, the late yellow flag caused some mayhem.

Re: The Great Big 2021 A > B > C Thread

Posted: Mon Apr 19, 2021 11:08 pm
by Johnson
Looks like all the drivers already settled at teams are doing better than predicted result, except Verstappen.

Re: The Great Big 2021 A > B > C Thread

Posted: Sat May 01, 2021 10:33 pm
by Exediron
Portimao update:

NOTE: A number of people went quicker in Q2 than in Q3. I am aware of this, but am sticking to my methodology as described above when this issue came up in Bahrain.

Leclerc / SAINZ
Projected: 0.386% to 0.542% (Leclerc)
Current Result: 0.519% (Leclerc)
Data (click to show)

VERSTAPPEN / PEREZ
Projected: 0.385% - 0.544% (Verstappen)
Current Result: 0.182% (Verstappen)
Data (click to show)

NORRIS / RICCIARDO
Projected: 0.262% - 0.403% (Ricciardo)
Current Result: 0.052% (Ricciardo)
Data (click to show)

STROLL / VETTEL
Projected: 0.262% - 0.316% (Vettel)
Current Result: 0.339% (Stroll)
Data (click to show)

OCON / ALONSO
Projected: 0.158% - 0.426% (Alonso)
Current Result: 0.629% (Ocon)
Data (click to show)

With a third data point added to each match-up, we're now switching from the mean to median average. This had minimal effect on several pairings: LEC / SAI, NOR / RIC, and VER / PER all changed only slightly, since either the first two results were very similar (Ferrari, McLaren) or the new data point fell neatly into the middle of the two (Red Bull).

How are we doing relative to projections so far?

Well, that depends on who you're looking at. The Ferrari pairing is still tracking within the projection, although Sainz has managed to dent Leclerc's run of results. At Red Bull, Verstappen has not been able to execute well enough to deliver on his projected advantage over Perez in qualifying, but the pace seems to be there if he can just nail his Q3 laps at some point again.

And then we have three that are sharply against projections. In the NOR / RIC pairing, the current number is slightly below Ricciardo's projected advantage -- but the reality of the season has been utter dominance for Norris which is not reflected in the current average. If Dan is unable to reverse the trend, this one is going to end as an embarrassing defeat of the model.

Another pairing that is headed for an upset is at Alpine. Alonso's qualifying victory in Bahrain is looking more and more like a fluke as Ocon has now outpaced him decisively in two consecutive sessions since. The race pace appears much more even, but in terms of one lap speed Alonso clearly has quite a bit to find.

And finally, the Aston Martin pairing. This one started out with an upset, but the trend has been rapidly heading in Vettel's favor, and we may well have already seen this match-up correct to within the projection.

Re: The Great Big 2021 A > B > C Thread

Posted: Sun May 02, 2021 11:56 am
by pokerman
I can't get my head around basically ignoring the 1 second gap Norris just pulled on Ricciardo to have Ricciardo ahead.

Also the gap Vettel pulled on Stroll gets ignored so instead of having it level we have Stroll confortably ahead.

Re: The Great Big 2021 A > B > C Thread

Posted: Sun May 02, 2021 9:58 pm
by Exediron
pokerman wrote:
Sun May 02, 2021 11:56 am
I can't get my head around basically ignoring the 1 second gap Norris just pulled on Ricciardo to have Ricciardo ahead.

Also the gap Vettel pulled on Stroll gets ignored so instead of having it level we have Stroll confortably ahead.
It'll all even out at season's end. You aren't going to try to tell me you think Norris' gap was actually representative, are you?

Allowing huge gaps from a single session to pollute the data isn't better in my opinion. Yes, the Norris / Ricciardo battle is currently misleading, but that's because of Ricciardo managing to qualify ahead in the first two rounds despite being slower, not because of blunting the impact of Norris' ridiculous qualifying margin in Portugal.

Re: The Great Big 2021 A > B > C Thread

Posted: Mon May 03, 2021 12:19 am
by pokerman
Exediron wrote:
Sun May 02, 2021 9:58 pm
pokerman wrote:
Sun May 02, 2021 11:56 am
I can't get my head around basically ignoring the 1 second gap Norris just pulled on Ricciardo to have Ricciardo ahead.

Also the gap Vettel pulled on Stroll gets ignored so instead of having it level we have Stroll confortably ahead.
It'll all even out at season's end. You aren't going to try to tell me you think Norris' gap was actually representative, are you?

Allowing huge gaps from a single session to pollute the data isn't better in my opinion. Yes, the Norris / Ricciardo battle is currently misleading, but that's because of Ricciardo managing to qualify ahead in the first two rounds despite being slower, not because of blunting the impact of Norris' ridiculous qualifying margin in Portugal.
I actually think it is, it gives him an average of about 0.3s on Ricciardo which is about the true gap, Norris has had that kind of performance advantage on Ricciardo but messed up his first 2 qualifying sessions, Ricciardo actually didn't even mess up his session, he really was that slow, at this stage having Ricciardo as the faster driver seems totally false.

Its horses for courses I guess, we have our different systems, you have your mathematical system whereas I'm more hands on, I look to sort the wheat from the chaff, I look to post some numbers after Barcelona, provided it doesn't rain, I like to get a feel for how the drivers are actually performing which takes a few races.

Re: The Great Big 2021 A > B > C Thread

Posted: Mon May 03, 2021 1:08 am
by Exediron
pokerman wrote:
Mon May 03, 2021 12:19 am
Exediron wrote:
Sun May 02, 2021 9:58 pm
pokerman wrote:
Sun May 02, 2021 11:56 am
I can't get my head around basically ignoring the 1 second gap Norris just pulled on Ricciardo to have Ricciardo ahead.

Also the gap Vettel pulled on Stroll gets ignored so instead of having it level we have Stroll confortably ahead.
It'll all even out at season's end. You aren't going to try to tell me you think Norris' gap was actually representative, are you?

Allowing huge gaps from a single session to pollute the data isn't better in my opinion. Yes, the Norris / Ricciardo battle is currently misleading, but that's because of Ricciardo managing to qualify ahead in the first two rounds despite being slower, not because of blunting the impact of Norris' ridiculous qualifying margin in Portugal.
I actually think it is, it gives him an average of about 0.3s on Ricciardo which is about the true gap, Norris has had that kind of performance advantage on Ricciardo but messed up his first 2 qualifying sessions, Ricciardo actually didn't even mess up his session, he really was that slow, at this stage having Ricciardo as the faster driver seems totally false.

Its horses for courses I guess, we have our different systems, you have your mathematical system whereas I'm more hands on, I look to sort the wheat from the chaff, I look to post some numbers after Barcelona, provided it doesn't rain, I like to get a feel for how the drivers are actually performing which takes a few races.
I should note that, although I'm posting these race-by-race, my system is only intended to give a representative result at season's end. The median system requires a fair amount of data fed into it before it can deliver a useful result.

That aside, I'm not objecting to the idea that ~0.3 would be an accurate gap for Norris -- I think it might be. I'm objecting to the idea of allowing a gap of over a second (garbage data, essentially) to contribute to a lasting total average.

In my method, at the end of the season it will just become an extreme 'win' for Norris, pushing the average one click further in the direction of a Norris win. In your system, it will be worth an extra 0.056% (assuming 23 data points) to Norris' season ending total regardless of how the rest of the season pans out. That's more than the difference between several drivers right there, all because of a single wide disparity in a single session early in the season.

Re: The Great Big 2021 A > B > C Thread

Posted: Mon May 03, 2021 11:04 am
by A.J.
A median is almost always a better data point to have than the mean (average) - just imagine if Ricciardo (hypothetically) beats Norris by a tenth in ever single qualifying session except 2, where Norris has a 1s advantage over him for whatever reason (such as in Portugal). The median will tell you that Ricciardo beat Norris by a tenth over the course of the season (which is accurate), while the average between them over a 20 race season will show Norris ahead by a bit, even though he was beaten 18 times out of 20 in this example.

Same reason why America has such issues with poverty, despite the mean income being amongst the highest in the world.

Re: The Great Big 2021 A > B > C Thread

Posted: Mon May 03, 2021 1:35 pm
by pokerman
I understand both points of view, however I believe that drivers are always making micro mistakes which allowed Ricciardo to out qualify Norris in the first 2 qualifiers despite Norris basically being faster, then say Ricciardo manages to pile all of those mistakes into one session creating an outlier and then the gap itself gets ignored, although in this case that was as fast as Ricciardo could go.

So in this case a 1 second gap gets ignored as an outlier nominally giving Ricciardo an advantage at this stage but the actual reality is that over the 3 sessions Norris has been 1 second quicker but Norris has lost all his early advantage.

Like I say it's each to his own, one is pure scientific methodology the other incorporating a touch of common-sense which I guess can be viewed as having too much personal influence on the outcome because I would be deciding what's admissible and what is not, for instance a 1 second gap were a driver as not really being able to set his best lap because of yellow flags then doesn't get thrown into the mix.

So looking at a post were one driver is always marginally quicker but then the other driver throws into the mix two 1 second gaps well that doesn't automatically get counted in my system for reasons already given another being wet sessions so basically anything that looks like garbage doesn't get used.

The more scientific method is just to use everything without context and let the mathematical construct sort it all out, it will be interesting to see what results the different methods come up with.

Re: The Great Big 2021 A > B > C Thread

Posted: Mon May 03, 2021 2:13 pm
by pokerman
Another point is that I would be guessing in the median system that whoever comes out on top in the head to head in qualifying then automatically comes out on top with the average gap so it's not really about averages as such?

Re: The Great Big 2021 A > B > C Thread

Posted: Mon May 03, 2021 10:27 pm
by Exediron
pokerman wrote:
Mon May 03, 2021 2:13 pm
Another point is that I would be guessing in the median system that whoever comes out on top in the head to head in qualifying then automatically comes out on top with the average gap so it's not really about averages as such?
That is correct. In the unlikely event that one driver was generally quicker than the other but lost the qualifying battle anyway, it will focus on the smallest advantage from the winning driver. Normally, this will be a ~0.050 or less result, indicating a nearly even comparison. I will note that in my time compiling these, I have not come across an instance where the quicker driver lost the qualifying head-to-head.

To the second part of your post: It is still an average -- it is simply a different sort of average. An average is about finding a number that expresses the middle of a data set, and mean and median are two different ways of arriving at that conclusion. Both are subject of potential flaws, but I prefer the median because it is much less sensitive to being influenced by abnormal data points. In a mean average you have to remove those data points entirely or allow them to dominate the data, whereas with a median you can keep them and they still express influence over the result without dominating it.