Page 1 of 9

The Great Big 2021 A > B > C Thread

Posted: Fri Dec 18, 2020 9:48 pm
by Exediron
As per this thread, I thought it would be interesting to compile an omnibus thread where we can agree on A > B > C predictions for all (or as many as possible) of the new 2021 driver pairings ahead of the season.

After the season, we can then come back and check those predictions and see how many were correct, how many were wrong, and by how much error.

NEW DRIVER PAIRINGS

Out of the ten Formula One teams, seven will have a different pairing from last season (the exceptions being Mercedes, Williams, and Alfa Romeo).

Leclerc / Sainz
Prediction: Leclerc by 0.386% to 0.542% (~0.30s to ~0.43s)
KingVoid: Leclerc by 0.30s
Pokerman: Leclerc by 0.47s
mikeyg123: Leclerc by 0.30s (guess)

The source for a cross comparison here is Leclerc / Vettel / Ricciardo, and then comparing Ricciardo to Max, Hulkenberg, and Kvyat, all of whom were also teammates with Sainz.

The main risk of this comparison is that Sainz has not had the same teammate for more than one complete season, apart from Norris. Kvyat was only his teammate for a partial season.


Verstappen / Perez
Prediction: Verstappen by 0.385% - 0.544% (~0.30s to ~0.43s)
KingVoid: Verstappen by 0.30s
Pokerman: Verstappen by 0.31s - 0.36s
mikeyg123: Verstappen by 0.25s (guess)

This prediction relies on the path of Verstappen / Ricciardo, and then comparing Ricciardo to Ocon and Hulkenberg, both of whom were also teammates with Perez.

Both links are single season on Ricciardo's end, but Perez and Hulkenberg have a very solid connection as do Verstappen and Ricciardo.


Norris / Ricciardo
Prediction: Ricciardo by 0.262% - 0.403% (~0.20s to ~0.32s)
KingVoid: Ricciardo by 0.12s
Pokerman: Ricciardo by 0.21s - 0.26s.
mikeyg123: Ricciardo by 0.15s (guess)

Since Sainz is Norris' only teammate, this comparison has to rely on Sainz. Sainz / Verstappen / Ricciardo is one avenue, and also Sainz / Hulkenberg / Ricciardo.

Sainz and Norris have been teammates for two seasons, so there is a decent prediction value to comparing Sainz.


Stroll / Vettel
Prediction: Vettel by 0.262% - 0.316% (~0.20s to ~0.25s)
KingVoid: Vettel by 0.25s
Pokerman: Vettel by 0.20s
mikeyg123: Vettel by 0.25s (guess)

This one is more difficult, because the link between Vettel and Stroll is not very direct. Stroll has two possible comparisons: Massa, and Perez. Vettel can be linked to both of them.

For Massa, Vettel / Raikkonen = Massa / Stroll is the path. For Perez, Vettel / Ricciardo / Ocon / Perez is the most direct. For a more byzantine cross comparison, Vettel / Raikkonen / Alonso / Button / Perez / Stroll is also a data point.


Ocon / Alonso
Prediction: Alonso by 0.158% - 0.426% (~0.12s to ~0.34s)
KingVoid: Alonso by less than 0.24s
Pokerman: Alonso by 0.15s
mikeyg123: Alonso by 0.45s (guess)

Ocon has been teammates with Perez and Ricciardo; of the two, Perez has the better link to Alonso. The previously mentioned indirect link of Alonso / Button / Perez / Ocon works here, with the advantage that both Alonso / Button and Perez / Ocon were two-season pairings.

The other angle is Alonso / Raikkonen / Vettel / Ricciardo / Ocon.


Gasly / Tsunoda
Prediction: 0.00 - 0.00
Pokerman: Gasly by 0.10s (guess)

This prediction is not possible using F1 data, as Tsunoda is a rookie. If someone can find an F2 link, I'd be open to hearing it!

Schumacher / Mazepin
Prediction: 0.00 - 0.00
Pokerman: Schumacher by 0.10s (guess)

Both drivers are rookies, so any common link would have to be found in the junior categories.

---

My concept here is to explore as many different cross comparisons as possible in an effort to establish a range for each pairing.

Posters in the thread are welcome to contribute in any way they want, but if you simply don't believe in cross comparisons and think they're stupid that isn't actually a contribution. ;)

Updated 12/18/2020 with KingVoid's numbers
Updated 12/19/2020 with Pokerman's numbers (and two guesses)
Updated 12/20/2020 to include my prediction on Leclerc / Sainz
Updated 12/21/2020 to include my prediction on Verstappen / Perez
Updated 12/23/2020 to include my prediction on Ricciardo / Norris
Updated 12/24/2020 to include my prediction on Alonso / Ocon
Updated 12/27/2020 to expand the range for Alonso / Ocon
Updated 12/28/2020 to include my prediction on Vettel / Stroll
Updated 12/28/2020 to formally add in mikey's guesses

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

Posted: Fri Dec 18, 2020 9:49 pm
by Exediron
Having created this thread, I don't have time to start populating it now. Optimistically, later tonight I hope to begin!

If anyone wants to chime in before then, feel free. I know pokerman has extensive numbers on most of these already.

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

Posted: Fri Dec 18, 2020 9:58 pm
by mikeyg123
I don't have any data to contribute but I'd just like to say before anyone chimes in about how A > B > C comparisons are useless and unreliable that whilst they can be wrong they are far more often right especially if you remove debut seasons and retirement years. Most of the anomalies can be explained by one of those featuring.

It's certainly by far the most realistic modelling tool for comparing drivers and making predictions we have.

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

Posted: Fri Dec 18, 2020 10:13 pm
by Invade
Crude, imprecise, but generally right from what I hear.

I'd like to know how often these sorts of comparisons have been used in the past to guesstimate the qualifying performances between new pairings and the success rate for at least picking the right winner, especially in cases where such a model supposes a very close edge of a tenth or so.

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

Posted: Fri Dec 18, 2020 10:22 pm
by mikeyg123
Ok I'll give some guesses without any actually statistics to hand just for fun.

In my cross comparison I will attempt to avoid rookie seasons or seasons where a driver new early on it was going to be there last in F1.

Leclerc/Sainz

I'm going Leclerc - Vettel - Ricciardo - Hulk - Sainz.

I believe this puts Sainz at a similar level to Vettel so I am going to say the gap will be Leclerc 0.3 SECONDS AHEAD

Verstappen/Perez

Again the easiest route is using Ricciardo. I'm going Verstappen - Ricciardo - Hulkenberg - Perez.

Basically each driver was very slightly slower than the one further up the chain so I am going with VERSTAPPEN 0.25 AHEAD

Norris/Ricciardo

The only reasonable route Norris - Sainz - Hulkenberg - Ricciardo.

This one should be fairly close. Norris, Sainz and Hulk all had roughly the same pace. Hulk with slightly more and Ricciardo with slightly more than him. With that in mind my guess is RICCIARDO 0.15 SECONDS AHEAD.

Stroll/Vettel

The fastest way to get to them is through Massa and Kimi but that would mean using Stroll's rookie season. However we can go one way without having to use Stroll's rookie season so that would be Stroll - Perez - Hulkenberg - Ricciardo - Vettel.

Interestingly this comparison put's Vettel and Perez at a similar level. I'm going with VETTEL 0.25 SECONDS AHEAD.

Ocon/Alonso

My chosen way of connecting them is Ocon - Ricciardo - Vettel - Raikkonen - Alonso. This is possibly the hardest cross comparison but I would suggest it puts ALONSO 0.45 SECONDS AHEAD.

Interesting how the same small pool of drivers keep coming up for all of them. Ricciardo, Sainz, Perez, Hulk and Vettel. Ricciardo features in every single one.

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

Posted: Fri Dec 18, 2020 10:23 pm
by mikeyg123
Invade wrote:
Fri Dec 18, 2020 10:13 pm
Crude, imprecise, but generally right from what I hear.

I'd like to know how often these sorts of comparisons have been used in the past to guesstimate the qualifying performances between new pairings and the success rate for at least picking the right winner, especially in cases where such a model supposes a very close edge of a tenth or so.
I think anything under a tenth and it gets harder. Especially in a chain with many links. Luckily it is actually quite rare for team mates to be that close.

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

Posted: Fri Dec 18, 2020 10:24 pm
by FrogInARaceCar
A request for those who've built the comparison models: Provide your confidence intervals. You're never going to get it bang on, but something like "I think Vettel will beat Stroll by 0.3s, plus or minus 0.2s", gives us all a much better idea of how well you think your method works and gives a way to score hits and misses.

Also, bear in mind that the tighter your confidence margins, the better you look when you're right but the worse when you're wrong...

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

Posted: Fri Dec 18, 2020 11:05 pm
by Exediron
mikeyg123 wrote:
Fri Dec 18, 2020 10:22 pm
Interesting how the same small pool of drivers keep coming up for all of them. Ricciardo, Sainz, Perez, Hulk and Vettel. Ricciardo features in every single one.
I noticed that as well when I was writing the suggestions! Ricciardo leaving Red Bull, and partnering with drivers who have had several different teammates themselves, has been wonderful for cross comparisons.
FrogInARaceCar wrote:
Fri Dec 18, 2020 10:24 pm
A request for those who've built the comparison models: Provide your confidence intervals. You're never going to get it bang on, but something like "I think Vettel will beat Stroll by 0.3s, plus or minus 0.2s", gives us all a much better idea of how well you think your method works and gives a way to score hits and misses.

Also, bear in mind that the tighter your confidence margins, the better you look when you're right but the worse when you're wrong...
I built the thread to include a range partly for that reason, but I will certainly add something about my own confidence when I add data to it. As a general rule of thumb for predicting the winner of the pairing I believe these comparisons to be highly accurate (90%+) when the measured gap is two tenths or above, generally accurate (maybe 60%-70%) when the gap is about a tenth, and not accurate when it's much under a tenth.

That would imply that my personal confidence in the numbers I've seen is about +/- 0.10 of a second.

I should also add as an addendum that I think figures which use percentage are more accurate than those measuring in pure time, since some tracks are much longer than others. If Leclerc out-qualifies Vettel by 0.250 at Monaco and is then out-qualified in return by 0.250 at Singapore, those should not balance each other out.

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

Posted: Fri Dec 18, 2020 11:24 pm
by JN23
mikeyg123 wrote:
Fri Dec 18, 2020 10:22 pm
Ocon/Alonso

My chosen way of connecting them is Ocon - Ricciardo - Vettel - Raikkonen - Alonso. This is possibly the hardest cross comparison but I would suggest it puts ALONSO 0.45 SECONDS AHEAD.
Nice post Mikey, thanks for sharing. Thanks to Exediron too, nice idea for a thread.

Question on this, if you replaced Ocon with Verstappen you should get a comparison between Verstappen and Alonso using the same route. What would you make that?

If you assume Perez = Ocon (roughly and for arguments sake) and Verstappen is predicted to be 0.25 faster than Perez, does that make Alonso 0.2 faster than Verstappen?

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

Posted: Fri Dec 18, 2020 11:28 pm
by JN23
A question for those who keep data on gaps between teammates. If you have a pair of teammates A & B. If A gets knocked out in Q1 and B goes into Q2, do you use their times in Q1 where they both set a time to compare or do use B’s Q2 time compared to A’s Q1 time.

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

Posted: Fri Dec 18, 2020 11:33 pm
by Exediron
JN23 wrote:
Fri Dec 18, 2020 11:28 pm
A question for those who keep data on gaps between teammates. If you have a pair of teammates A & B. If A gets knocked out in Q1 and B goes into Q2, do you use their times in Q1 where they both set a time to compare or do use B’s Q2 time compared to A’s Q1 time.
I use the last session they set comparable times in, whatever that was -- so if one teammate makes it to Q2 and the other one doesn't, I use Q1. If both make it to Q2 and one driver's engine fails before he sets a lap, I use Q1.

It's not perfect, because it's entirely possible that a driver would set their best lap later in the session, but due to track evolution the laps aren't entirely comparable unless both drivers were on track at the same time.

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

Posted: Fri Dec 18, 2020 11:54 pm
by KingVoid
Leclerc ahead of Sainz by 0.30s

Verstappen ahead of Perez by 0.30s

Ricciardo ahead of Norris by 0.12s

Vettel ahead of Stroll by 0.25s

I cannot predict Alonso vs Ocon, because I don’t think that Alonso will be at his peak anymore. He won’t have declined as much as Schumacher, but he will still have declined.

That said, prime Alonso has around 0.24s on Ocon

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

Posted: Sat Dec 19, 2020 12:13 am
by KingVoid
The fact that Red Bull have signed Perez is great news for cross comparison merchants like myself and Poker.

We now have a much closer chain from Hamilton to Verstappen

Hamilton-Button-Perez-Verstappen

Button and Perez were dead even in 2013 on single lap pace, so there is only one question left: can Max beat Checo by a bigger margin than Lewis had over JB?

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

Posted: Sat Dec 19, 2020 2:56 am
by Exediron
KingVoid wrote:
Fri Dec 18, 2020 11:54 pm
Leclerc ahead of Sainz by 0.30s

Verstappen ahead of Perez by 0.30s
As a tangential point, wouldn't this suggest that Leclerc is slightly ahead of Verstappen, as suggested in the other thread? Perez was slightly slower than Hulk in qualifying, and Hulk was in turn slightly quicker than Sainz. If Verstappen is quicker than Leclerc, he ought to have a larger advantage over Perez than Leclerc has over Sainz.

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

Posted: Sat Dec 19, 2020 4:58 am
by Exediron
I've worked up Vettel / Ricciardo (2014) and Vettel / Leclerc (2019, 2020) so far, and I'm taking a break to let my brain recover. I'll go over my own methodology and results, to give anyone time to challenge any of the samples I've included or thrown out.

Methodology

I'm deviating from what seems to be the P-F1 standard in two ways: I'm using percentages instead of raw times gaps, and I'm using the median average instead of the mean. In both cases I believe my way paints a more accurate picture and does a better job reducing the impact of outliers.

Because of the reduced impact of statistical outliers with the median average, I am including wet qualifying sessions. I am not including sessions where one driver or the other was unable to set at least one comparable lap (this has come up twice so far, both times in the 2019 season).

Results So Far

Ricciardo > Vettel: 0.233%
Data (click to show)
This comparison came out with a larger margin for Ricciardo than I usually see. I don't know if I forgot to throw out some data point in his favor, or if it's just a difference of the percent/median method.

Leclerc > Vettel: (2019) 0.188%
Data (click to show)
Leclerc > Vettel: (2020) 0.423%
Data (click to show)
Leclerc > Vettel: (2019 + 2020) 0.267%
(2019 has more weight because there were more races in 2019)

Obviously, the inclusion or lack thereof for 2019 makes a huge difference to any comparison involving Leclerc. I expect a similar effect when I look at Verstappen; 2016 was a very different year for him (also his second in the sport) compared to 2017 and 2018.

Tomorrow (or maybe later tonight, after my batteries recharge) I intend to start working up Ricciardo against all his other teammates.

EDITED to correct the final Leclerc > Vettel (2019 + 2020) number.

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

Posted: Sat Dec 19, 2020 5:18 am
by KingVoid
Exediron wrote:
Sat Dec 19, 2020 2:56 am
KingVoid wrote:
Fri Dec 18, 2020 11:54 pm
Leclerc ahead of Sainz by 0.30s

Verstappen ahead of Perez by 0.30s
As a tangential point, wouldn't this suggest that Leclerc is slightly ahead of Verstappen, as suggested in the other thread? Perez was slightly slower than Hulk in qualifying, and Hulk was in turn slightly quicker than Sainz. If Verstappen is quicker than Leclerc, he ought to have a larger advantage over Perez than Leclerc has over Sainz.
I believe that Hulkenberg’s gap to Sainz was slightly larger than his gap to Perez, which would imply that Perez is slightly quicker than Sainz.

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

Posted: Sat Dec 19, 2020 9:14 am
by JN23
Exediron wrote:
Fri Dec 18, 2020 11:33 pm
JN23 wrote:
Fri Dec 18, 2020 11:28 pm
A question for those who keep data on gaps between teammates. If you have a pair of teammates A & B. If A gets knocked out in Q1 and B goes into Q2, do you use their times in Q1 where they both set a time to compare or do use B’s Q2 time compared to A’s Q1 time.
I use the last session they set comparable times in, whatever that was -- so if one teammate makes it to Q2 and the other one doesn't, I use Q1. If both make it to Q2 and one driver's engine fails before he sets a lap, I use Q1.

It's not perfect, because it's entirely possible that a driver would set their best lap later in the session, but due to track evolution the laps aren't entirely comparable unless both drivers were on track at the same time.
Thanks :thumbup: it does make sense this way but wanted to check.

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

Posted: Sat Dec 19, 2020 11:01 am
by JN23
KingVoid wrote:
Sat Dec 19, 2020 12:13 am
The fact that Red Bull have signed Perez is great news for cross comparison merchants like myself and Poker.

We now have a much closer chain from Hamilton to Verstappen

Hamilton-Button-Perez-Verstappen

Button and Perez were dead even in 2013 on single lap pace, so there is only one question left: can Max beat Checo by a bigger margin than Lewis had over JB?
What do you make Hamilton vs Button? I make it 0.21 in Hamilton's favour which is smaller than I'd have expected.

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

Posted: Sat Dec 19, 2020 1:12 pm
by pokerman
Exediron wrote:
Fri Dec 18, 2020 9:49 pm
Having created this thread, I don't have time to start populating it now. Optimistically, later tonight I hope to begin!

If anyone wants to chime in before then, feel free. I know pokerman has extensive numbers on most of these already.
Yes I will give my thoughts, presently watching Liverpool beating Palace. :)

I did think about starting a prediction thread, this would be a better version of what I was thinking. :thumbup:

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

Posted: Sat Dec 19, 2020 1:17 pm
by pokerman
mikeyg123 wrote:
Fri Dec 18, 2020 9:58 pm
I don't have any data to contribute but I'd just like to say before anyone chimes in about how A > B > C comparisons are useless and unreliable that whilst they can be wrong they are far more often right especially if you remove debut seasons and retirement years. Most of the anomalies can be explained by one of those featuring.

It's certainly by far the most realistic modelling tool for comparing drivers and making predictions we have.
Indeed also past performance has to be a guide to future performance otherwise how would you be able to predict anything in the first place beyond stick a tail on the donkey. :)

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

Posted: Sat Dec 19, 2020 1:37 pm
by pokerman
Invade wrote:
Fri Dec 18, 2020 10:13 pm
Crude, imprecise, but generally right from what I hear.

I'd like to know how often these sorts of comparisons have been used in the past to guesstimate the qualifying performances between new pairings and the success rate for at least picking the right winner, especially in cases where such a model supposes a very close edge of a tenth or so.
In recent years I would say Hamilton beating Bottas, Alonso beating Kimi, and Ricciardo beating the Hulk and Ocon were all predictable, I can't think of any surprises.

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

Posted: Sat Dec 19, 2020 1:59 pm
by FrogInARaceCar
I'll say what I expect from a qualitative point of view:

Leclerc to beat Sainz, but Sainz to be better than Vettel was.
Vettel to have roughly Perez's margin over Stroll.

The questions behind both of these is really "What was going on at Ferrari in 2020 - was Leclerc really good or was Vettel really bad, or was it a bit of both? If it was a Vettel problem, will he recover in 2021?"

Verstappen to be generally, but not always, ahead of Perez.

Ricciardo to be noticably better than Norris.

Alonso to comfortably beat Ocon.

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

Posted: Sat Dec 19, 2020 2:03 pm
by pokerman
JN23 wrote:
Fri Dec 18, 2020 11:24 pm
mikeyg123 wrote:
Fri Dec 18, 2020 10:22 pm
Ocon/Alonso

My chosen way of connecting them is Ocon - Ricciardo - Vettel - Raikkonen - Alonso. This is possibly the hardest cross comparison but I would suggest it puts ALONSO 0.45 SECONDS AHEAD.
Nice post Mikey, thanks for sharing. Thanks to Exediron too, nice idea for a thread.

Question on this, if you replaced Ocon with Verstappen you should get a comparison between Verstappen and Alonso using the same route. What would you make that?

If you assume Perez = Ocon (roughly and for arguments sake) and Verstappen is predicted to be 0.25 faster than Perez, does that make Alonso 0.2 faster than Verstappen?
Yes Alonso beating Ocon by 0.45s has to be seen as being well wide of the mark.

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

Posted: Sat Dec 19, 2020 2:06 pm
by pokerman
Exediron wrote:
Fri Dec 18, 2020 11:33 pm
JN23 wrote:
Fri Dec 18, 2020 11:28 pm
A question for those who keep data on gaps between teammates. If you have a pair of teammates A & B. If A gets knocked out in Q1 and B goes into Q2, do you use their times in Q1 where they both set a time to compare or do use B’s Q2 time compared to A’s Q1 time.
I use the last session they set comparable times in, whatever that was -- so if one teammate makes it to Q2 and the other one doesn't, I use Q1. If both make it to Q2 and one driver's engine fails before he sets a lap, I use Q1.

It's not perfect, because it's entirely possible that a driver would set their best lap later in the session, but due to track evolution the laps aren't entirely comparable unless both drivers were on track at the same time.
This. :thumbup:

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

Posted: Sat Dec 19, 2020 2:10 pm
by pokerman
KingVoid wrote:
Fri Dec 18, 2020 11:54 pm
Leclerc ahead of Sainz by 0.30s

Verstappen ahead of Perez by 0.30s

Ricciardo ahead of Norris by 0.12s

Vettel ahead of Stroll by 0.25s

I cannot predict Alonso vs Ocon, because I don’t think that Alonso will be at his peak anymore. He won’t have declined as much as Schumacher, but he will still have declined.

That said, prime Alonso has around 0.24s on Ocon
Yeah difficult to predict for Alonso, I also wondering about having a slight variance for drivers joining new teams with the incumbent driver having an initial advantage.

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

Posted: Sat Dec 19, 2020 2:17 pm
by pokerman
KingVoid wrote:
Sat Dec 19, 2020 12:13 am
The fact that Red Bull have signed Perez is great news for cross comparison merchants like myself and Poker.

We now have a much closer chain from Hamilton to Verstappen

Hamilton-Button-Perez-Verstappen

Button and Perez were dead even in 2013 on single lap pace, so there is only one question left: can Max beat Checo by a bigger margin than Lewis had over JB?
Well I had Button slightly up on Perez however I unfortunately will be guessing yes.

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

Posted: Sat Dec 19, 2020 9:05 pm
by Exediron
Continuing onward with calculating the prior teammate battles using my methodology, I have finished with Verstappen / Ricciardo. Out of the ones I've done so far, this comparison I would say differs most dramatically from the norms I usually see.

I have the teammate battle following a much steadier slope than the mean average method does; in 2016 I have Ricciardo firmly ahead, and in 2018 I have Verstappen firmly ahead, but in 2017 I have the two close to equal, which brings the entire comparison significantly towards the middle compared to most numbers.

This is caused almost entirely by dulling the effect of a single dry qualifying outlier from 2017, namely the Mexican Grand Prix. Ricciardo set a time almost a whole second slower than Max, and since there's no reason to throw the data out (no engine failure, not a wet session) it usually gets included, and heavily skews Max's season-long result upward.

Here are my results:

Ricciardo > Verstappen: (2016) 0.175%
Data (click to show)
I'm a little suspicious of Verstappen's qualifying time from Baku, being a massive 1.5% off Ricciardo. It's possible it should be thrown out if Max was having problems of some sort. Doing that would shift the median solidly onto the 0.144% line.

Verstappen > Ricciardo: (2017) 0.043%
Data (click to show)
Here we see the largest difference between the median and mean average methods. Verstappen's strong results in 2017 generated larger differences than Ricciardo's strong results, but the median between them is actually on the Verstappen-leaning end of the patch of sub-one tenth differences in the middle.

Verstappen > Ricciardo: (2018) 0.211%
Data (click to show)
Here, the trends converge again. Verstappen's advantage over Ricciardo is sustained enough that it pushes the median into the point where we see a solid, but not overwhelming, advantage for Max.

Now, the whole-partnership analysis.

This varies dramatically by what you allow into the sample. Taking all three seasons together, the conclusion is that Max has a very small margin over Ricciardo:

Verstappen > Ricciardo: (2016-2018) 0.054%
Data (click to show)
But I actually think it's unfair to include 2016 for Max, since he changed teams mid-season and was obviously off the pace compared to his new teammate for the first 1/3 or so.

Using only 2017 and 2018, we get a picture that's certainly more favorable to Max, although still a little shy of his 2018 mark. Considering 2018 is muddied a bit by Ricciardo becoming discontent with the team and Max assuming clear #1 status, I think the 2017-2018 range is probably the most accurate.

Verstappen > Ricciardo: (2017-2018) 0.181%
Data (click to show)
Next up, I'll look at Ricciardo's teammates at Renault: Hulkenberg and Ocon.

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

Posted: Sat Dec 19, 2020 9:17 pm
by pokerman
Right here goes, looking at Leclerc there's not many data points for him, this year he was either excellent or Vettel was sub par, nevertheless at this moment in time I can only go with what happened so:-
Leclerc > Sainz 0.47s

Onto the Max v Perez matchup plenty of data points here to make a more valid judgement with a bit of variance for Perez settling into his new team, so:-
Max > Perez 0.31s - 0.36s

Next the Ricciardo/Norris matchup, we have very recent data for this again to make a solid predication, again a bit of variance for Ricciardo settling into his new team, so:-
Ricciardo > Norris 0.21s - 0.26s.

Moving onto Vettel and Stroll, if he wants to recover his reputation then really needs to improve on this prediction:-
Vettel > Stroll 0.2s

Alonso against Ocon, can Alonso be back at his best, he needs to because even at his best I would be predicting this:-
Alonso > Ocon 0.15s

Next is something we can't really know, Gasly against Tsunoda, I see Gasly as being a little better than average while Tsunoda is talented and a fast learner, I think he can contain Gasly to no more than this prediction:-
Gasly > Tsunoda 0.1s

Finally two rookies Schumacher and Mazepin, both slow learners so who has impressed me more as juniors and that would be Schumacher, however I don't see Schumacher as being super fast so I predict:-
Schumacher > Mazepin 0.1s

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

Posted: Sun Dec 20, 2020 7:57 am
by Exediron
Okay, I've finished up a few more using my own methodology: Ricciardo / Hulk, Ricciardo / Ocon, Hulk / Sainz, and also Hulk / Perez. This is enough to start making my own predictions, but for now here are the comparisons themselves:

Ricciardo > Hulkenberg: (2019) 0.171%
Data (click to show)
This was a rare care where the median actually gave a larger differential than the mean average. The reason for that is that when Hulk did out-qualify Ricciardo, it was often by a sizable margin -- but it didn't happen very often.

My model continues to be higher on Ricciardo than mean average cross-comparisons.

Ricciardo > Ocon: (2020) 0.231%
Data (click to show)
There's no data set or method you can use that would make this anything but one-sided. Ocon only out-qualified Ricciardo twice all year, and a sizable number of Ricciardo's results ahead were large gaps.

And now for the Hulkenberg comparisons (that don't include Ricciardo):

Hulkenberg > Sainz: (2020) 0.181%
Data (click to show)
I'm on record as saying I think this comparison is slanted too heavily towards Hulk, but digging into the numbers I'm less certain. Yes, Sainz did begin the year poorly with four straight defeats -- but he also ended the season the same way. At any rate, the conclusion is that Hulk firmly shaded Sainz during their season together.

This is particularly interesting, considering...

Hulk and Perez!


Now, this one is a little weird, in much the same way Ricciardo vs. Verstappen is. The first year is very different from the other two. In 2014, Hulk was decidedly quicker than Perez (although this is somewhat affected by an abnormally large number of wet races, nearly all of which had Hulk faster). In 2015 and 2016, however, the result was very even -- essentially equal.

Hulkenberg > Perez: (2014) 0.186%
Data (click to show)
Hulkenberg > Perez: (2015) 0.032%
Data (click to show)
Hulkenberg > Perez: (2016) 0.035%
Data (click to show)
For the combined number of Hulk and Perez, I've come up with two sets of numbers: one including all of 2014-2016, and another using only 2015 and 2016. The two aren't as different as one might imagine, but of course the 2014-2016 sample is less close:

Hulkenberg > Perez: (2014, 2015, 2016) 0.075%
Hulkenberg > Perez: (2015 & 2016) 0.033%

Either way, a very slight edge for Hulk.

Coming up next, the first of my predictions using the previously gathered numbers!

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

Posted: Sun Dec 20, 2020 8:13 am
by Exediron
INITIAL PREDICTION FOR Leclerc / SAINZ

And here is my first prediction, using the methodology above!

Leclerc / Vettel / Ricciardo / Hulkenberg / Sainz

Leclerc > Vettel: -0.267% to -0.423%
At the low end of the range is the average of both seasons; at the high end is using only 2020.

Ricciardo > Vettel: -0.233%

Which suggests:

Leclerc > Ricciardo: -0.034% to -0.190%

Now, bringing Hulkenberg and Sainz into it:

Ricciardo > Hulkenberg: -0.171%
Hulkenberg > Sainz: -0.181%

Suggesting:

Ricciardo > Sainz: -0.352%

And finally:

Leclerc > Sainz: -0.386% to -0.542%

Around a circuit with a roughly 1'20"s lap time, that gives a gap of 0.308 to 0.433 seconds. This is broadly in line with Kingvoid's and Pokerman's existing predictions.

Allowing for the fact that I consider the Sainz / Hulkenberg season to be a weak data point, and I also suspect the 2020 Vettel / Leclerc season is unusually skewed in Leclerc's direction, I would expect the smaller end of the range personally.

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

Posted: Sun Dec 20, 2020 2:35 pm
by pokerman
Exediron wrote:
Sun Dec 20, 2020 8:13 am
INITIAL PREDICTION FOR Leclerc / SAINZ

And here is my first prediction, using the methodology above!

Leclerc / Vettel / Ricciardo / Hulkenberg / Sainz

Leclerc > Vettel: -0.267% to -0.423%
At the low end of the range is the average of both seasons; at the high end is using only 2020.

Ricciardo > Vettel: -0.233%

Which suggests:

Leclerc > Ricciardo: -0.034% to -0.190%

Now, bringing Hulkenberg and Sainz into it:

Ricciardo > Hulkenberg: -0.171%
Hulkenberg > Sainz: -0.181%

Suggesting:

Ricciardo > Sainz: -0.352%

And finally:

Leclerc > Sainz: -0.386% to -0.542%

Around a circuit with a roughly 1'20"s lap time, that gives a gap of 0.308 to 0.433 seconds. This is broadly in line with Kingvoid's and Pokerman's existing predictions.

Allowing for the fact that I consider the Sainz / Hulkenberg season to be a weak data point, and I also suspect the 2020 Vettel / Leclerc season is unusually skewed in Leclerc's direction, I would expect the smaller end of the range personally.
Yes my figure is the uppermost number based on the 2020 season being 100% reliable.

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

Posted: Sun Dec 20, 2020 4:20 pm
by WHoff78
Would be interesting to know any of these numbers give any indication as to whether certain teams favour a lead driver or equal driver policy at all. I guess in reality even that world vary from season to season though depending on the team mate and the specific scenario for each season (whether a team are competing for the odd race win/podium, or in a close fight in the WCC with a few other teams)

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

Posted: Mon Dec 21, 2020 5:14 am
by Exediron
I've now done Perez / Ocon, which gives the final piece to estimate Verstappen / Perez (which I'll do in my next post, using both Ocon and Perez as data points).

Beginning this one, I expected it to be quite close -- and it is. Even in 2018, when Ocon out-qualified Perez by a decisive 16-5 margin, many of those sessions were fairly close. The two teammates were often found next to one another on the grid, and this parity is indicated in the results.

Perez > Ocon: (2017) 0.098%
Data (click to show)
Ocon > Perez: (2018) 0.128%
Data (click to show)
Putting both seasons together, the overall comparison favors Ocon slightly:

Ocon > Perez: (2017-2018) 0.032%
Data (click to show)

These numbers seem to bear out the commonly expressed shorthand that Perez = Ocon.

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

Posted: Mon Dec 21, 2020 5:29 am
by Exediron
INITIAL PREDICTION FOR VERSTAPPEN / PEREZ

Having crunched the requisite numbers, here is my prediction for Verstappen / Perez. This data is arrived at by comparing Verstappen / Ricciardo, and then looking at Ricciardo, Hulkenberg, Ocon, and Perez.

Verstappen > Ricciardo: -0.181%

I'm using the 2017 and 2018 only version, discarding 2016 for the comparison. Verstappen changed teams mid-season, and that unsurprisingly led to him starting his stint at Red Bull off the pace compared to Ricciardo.

Now, the comparison splits down two different avenues of cross-comparison:

Via Hulkenberg

Ricciardo > Hulkenberg: -0.171%
Hulkenberg > Perez: -0.033 - 0.075% (depending on if one considers 2014 an outlier)

Verstappen > Perez: -0.385 - 0.427%

Via Ocon

Ricciardo > Ocon: -0.231%
Ocon > Perez: -0.032 - 0.128% (using both seasons or only 2018)

Verstappen > Perez: -0.444 - 0.544%

Conclusion

The two comparisons produce a range, because they do not wholly agree. Ocon and Perez compared very similarly to Hulkenberg and Perez, but Ricciardo beat Ocon by a larger margin than he beat Hulkenberg. I've set the lower bound of the range by the Hulkenberg route, and the upper bound by the Ocon route.

Around a circuit with a roughly 1'20"s lap time, that gives a gap of 0.308 to 0.435 seconds. The keen-eyed will notice that this is almost precisely the same range I have estimated for Leclerc / Sainz, and broadly in line with Kingvoid's and Pokerman's predictions.

My personal interpretation is that Ricciardo was coming into the team against Hulkenberg, whereas against Ocon he was already established and thus had an advantage. I would therefore pick somewhere towards the middle of this range, roughly at the point where the two overlap, for my own prediction.

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

Posted: Mon Dec 21, 2020 1:20 pm
by pokerman
Exediron wrote:
Mon Dec 21, 2020 5:29 am
INITIAL PREDICTION FOR VERSTAPPEN / PEREZ

Having crunched the requisite numbers, here is my prediction for Verstappen / Perez. This data is arrived at by comparing Verstappen / Ricciardo, and then looking at Ricciardo, Hulkenberg, Ocon, and Perez.

Verstappen > Ricciardo: -0.181%

I'm using the 2017 and 2018 only version, discarding 2016 for the comparison. Verstappen changed teams mid-season, and that unsurprisingly led to him starting his stint at Red Bull off the pace compared to Ricciardo.

Now, the comparison splits down two different avenues of cross-comparison:

Via Hulkenberg

Ricciardo > Hulkenberg: -0.171%
Hulkenberg > Perez: -0.033 - 0.075% (depending on if one considers 2014 an outlier)

Verstappen > Perez: -0.385 - 0.427%

Via Ocon

Ricciardo > Ocon: -0.231%
Ocon > Perez: -0.032 - 0.128% (using both seasons or only 2018)

Verstappen > Perez: -0.444 - 0.544%

Conclusion

The two comparisons produce a range, because they do not wholly agree. Ocon and Perez compared very similarly to Hulkenberg and Perez, but Ricciardo beat Ocon by a larger margin than he beat Hulkenberg. I've set the lower bound of the range by the Hulkenberg route, and the upper bound by the Ocon route.

Around a circuit with a roughly 1'20"s lap time, that gives a gap of 0.308 to 0.435 seconds. The keen-eyed will notice that this is almost precisely the same range I have estimated for Leclerc / Sainz, and broadly in line with Kingvoid's and Pokerman's predictions.

My personal interpretation is that Ricciardo was coming into the team against Hulkenberg, whereas against Ocon he was already established and thus had an advantage. I would therefore pick somewhere towards the middle of this range, roughly at the point where the two overlap, for my own prediction.
Yeah that's why I've included a slight variance in a couple of my predictions, when you look at Renault it's like the incoming driver lost 0.05s as an average as he became accustomed to his new team.

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

Posted: Mon Dec 21, 2020 1:33 pm
by Covalent
These are qualifying times? I would be interested to see how the a-b-c comparisons worked out for past seasons, so prediction vs reality.

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

Posted: Mon Dec 21, 2020 2:50 pm
by pokerman
Covalent wrote:
Mon Dec 21, 2020 1:33 pm
These are qualifying times? I would be interested to see how the a-b-c comparisons worked out for past seasons, so prediction vs reality.
The most obvious one is Alonso/Massa/Kimi, Massa and Kimi were quite equal, Alonso beat Massa by about 3 tenths and then went on and beat Kimi by the same margin.

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

Posted: Wed Dec 23, 2020 8:09 am
by Exediron
I've skipped a few days, but I've not gotten around to doing the next on my list -- Norris / Sainz.

The general perception seems to be that the pair have been close, with Norris having the edge on raw pace but not race results. My personal findings suggest that it's Sainz with the edge, although it did decrease noticeably in Norris' second season.

Sainz > Norris: (2019) 0.118%
Data (click to show)
Sainz > Norris: (2020) 0.006%
Data (click to show)
Putting the two seasons together, we get an overall comparison that slightly favors Sainz, but basically comes up with them as quite evenly matched:

Sainz > Norris: (2019, 2020) 0.051%
Data (click to show)

I made one change from what might be the standard data set for this comparison, in Canada 2019. Sainz did set a time in Q3, which was about 2 seconds off what Norris set. I've decided to throw that out, and instead use the times from Q2 -- which had Sainz ahead.

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

Posted: Wed Dec 23, 2020 8:20 am
by Exediron
INITIAL PREDICTION FOR RICCIARDO / NORRIS

I'm comparing Ricciardo to Norris via the obvious path of Norris / Sainz / Hulkenberg / Ricciardo.

My prediction is going to be fairly harsh on Norris. My model has been consistently pretty high on Ricciardo, and since Sainz did quite poorly at Renault while Ricciardo was able to hammer Hulk quite effectively, the numbers are going to predict a very one-sided match-up here.

Ricciardo > Hulkenberg: -0.171%
Hulkenberg > Sainz: -0.181%

Which gives us:

Ricciardo > Sainz: -0.352%

And then a fairly small step up to:

Ricciardo > Norris: -0.403%

There's no built-in range here, since I'm only using a single data point at each step of the chain. Therefore, I'm going to assign my own range:

I think the calculated number is the upper end of the range. I've said before that I think Hulk / Sainz is a weak data point. Additionally, Norris is going into his third year, towards where I tend to believe is the end of a driver's outright speed development. I expect Norris will have taken another step forward compared to last season, equal to maybe 0.050-0.100%.

With that adjustment, and halving the impact of Hulk / Sainz to adjust for my dubiousness of the comparison, I'd give the lower bounds at 0.262%.

A range of 0.262% - 0.403%, around a circuit with a roughly 1'20"s lap time, would be equal to a gap of 0.209 to 0.322 seconds.

I am predicting a clear-cut dominance by Ricciardo, equal to or greater than the advantage he had over Ocon this season.

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

Posted: Wed Dec 23, 2020 1:13 pm
by pokerman
Exediron wrote:
Wed Dec 23, 2020 8:20 am
INITIAL PREDICTION FOR RICCIARDO / NORRIS

I'm comparing Ricciardo to Norris via the obvious path of Norris / Sainz / Hulkenberg / Ricciardo.

My prediction is going to be fairly harsh on Norris. My model has been consistently pretty high on Ricciardo, and since Sainz did quite poorly at Renault while Ricciardo was able to hammer Hulk quite effectively, the numbers are going to predict a very one-sided match-up here.

Ricciardo > Hulkenberg: -0.171%
Hulkenberg > Sainz: -0.181%

Which gives us:

Ricciardo > Sainz: -0.352%

And then a fairly small step up to:

Ricciardo > Norris: -0.403%

There's no built-in range here, since I'm only using a single data point at each step of the chain. Therefore, I'm going to assign my own range:

I think the calculated number is the upper end of the range. I've said before that I think Hulk / Sainz is a weak data point. Additionally, Norris is going into his third year, towards where I tend to believe is the end of a driver's outright speed development. I expect Norris will have taken another step forward compared to last season, equal to maybe 0.050-0.100%.

With that adjustment, and halving the impact of Hulk / Sainz to adjust for my dubiousness of the comparison, I'd give the lower bounds at 0.262%.

A range of 0.262% - 0.403%, around a circuit with a roughly 1'20"s lap time, would be equal to a gap of 0.209 to 0.322 seconds.

I am predicting a clear-cut dominance by Ricciardo, equal to or greater than the advantage he had over Ocon this season.
That would be the same as my prediction then. :)