[stat]Team mates comparison 2008[/stat]

[Schummy]

Here is the stat I did (incompletely) last seasons. For each race I assign three 1/0 values to every team. Values 1 means 1st driver beats 2nd driver, 0 means the reverse. First number is for grid position, second is for fastest race lap and the last one refers to race classification.

TEAM    AUS MAL BAH
=========================
RAI-MAS 011 011 000 = 4-5
HAM-KOV 101 010 100 = 4-5
HEI-KUB 011 010 010 = 4-5
ROS-NAK 111 111 111 = 9-0 <--
TRU-GLO 110 111 111 = 8-1 <--
ALO-PIQ 111 111 101 = 8-1 <--
WEB-COU 000 101 111 = 5-4
VET-BOU 100 111 000 = 4-5
BUT-BAR 000 111 100 = 4-5
FIS-SUT 100 111 111 = 7-2 <--
SAT-DAV 111 100 000 = 4-5
=========================

Plenty of equality in most of teams. Only Rosberg, Trulli, Alonso and Fisichella are overwhelming their team mates.

Interestingly, in many teams, the supposed 2nd driver is resisting or surpassing the first one.

Note: first-second driver position is subjectively chosen be me in terms of perceived a priori driving strength or performance. Most of cases are very clear, others are not, I'm afraid.

[Chiefy]

And do I get straight through to Nick Heidfeld if I dial that number?

[brands]

I feel the fastest race lap is misleading as many a time a driver will put in a really fast lap after 40-50 average ones. Surely a better test of speed would be the Q1/Q2 times ?

[Schummy]

Originally Posted by brands

I feel the fastest race lap is misleading as many a time a driver will put in a really fast lap after 40-50 average ones. Surely a better test of speed would be the Q1/Q2 times ?

I don't know if with Q1, Q2 you mean quartile or Qualification parts. If it is the second, qualification is already used in the table. I want to express race laps, what is different from qualifying laps (for example, Senna-Prost case), the simplest way is to use fastest laps, other choice is to use some other measure as median, quartile, etc. Average lap time is just equivalent to final race time (if the driver gets to the end) i.e. final classification (that is used in the table as well).

[Greenback]

Originally Posted by brands

I feel the fastest race lap is misleading as many a time a driver will put in a really fast lap after 40-50 average ones. Surely a better test of speed would be the Q1/Q2 times ?

I think the average of Q2 and fastest lap timing will be a better indicator.

[FPV GTHO]

Fastest race laps has its place. Most times its set at the end of a stint with worn tyres and low fuel. Some drivers are going to wear their tyres more than others, that doesnt get reflected in qualifying tmes. Maybe averaging the fastest 3 or 5 race laps would be more accurate for race pace though.

[Greenback]

Originally Posted by FPV GTHO

Fastest race laps has its place. Most times its set at the end of a stint with worn tyres and low fuel. Some drivers are going to wear their tyres more than others, that doesnt get reflected in qualifying tmes. Maybe averaging the fastest 3 or 5 race laps would be more accurate for race pace though.

If tyrewear become a variable, then how can the result be more accurate?

[FPV GTHO]

Managing the tyres is a critical part of being fast during the race.

[Greenback]

Originally Posted by FPV GTHO

Managing the tyres is a critical part of being fast during the race.

I would have thought that managing tyres is critical part of winning and finishing the race and not setting the fastest lap time. ie. if a driver have to nurse his tyres because of technical reasons, his ability to set the fastest lap would have been compromised. On the otherhand, suppose if his team mate is only concern about setting quick lap time at the expense of flat spotting his tyres etc then I am sure his lap time will be more impressive.

Furthermore, there are also many other factors that will affect a drivers ability to set the fastest lap.

tyres strategy
fuel strategy
weather and temperature
traffic
car setup (engineering problems)

[Schummy]

But, in a sense, managing tyres has an important input in setting absolute fastest laps. Prost has an excellent result in race fastest laps, much better than Senna (the reverse for poles). When the fuel load is low, if you has not cared the tyres you are not in a good position to exploit it. Somehow, drivers with good qualifying records have not automatically good race fast laps records.

Possibly this topic is an interesting subject for another stat

[Greenback]

Originally Posted by Schummy

But, in a sense, managing tyres has an important input in setting absolute fastest laps. Prost has an excellent result in race fastest laps, much better than Senna (the reverse for poles). When the fuel load is low, if you has not cared the tyres you are not in a good position to exploit it. Somehow, drivers with good qualifying records have not automatically good race fast laps records.

Possibly this topic is an interesting subject for another stat

thats the reason why I am suggesting that we should consider using the average of both fastest lap (during race) and Q2 timing.

[Schummy]

Originally Posted by Greenback

thats the reason why I am suggesting that we should consider using the average of both fastest lap (during race) and Q2 timing.

I don't fully understand your proposal. This thread is about this stat about team mates comparisons, and within it we have already fastest lap AND grid position. So I think I read your idea as substituting grid position with Q2.

[Greenback]

Originally Posted by Schummy

I don't fully understand your proposal. This thread is about this stat about team mates comparisons, and within it we have already fastest lap AND grid position. So I think I read your idea as substituting grid position with Q2.

Fuel load strategy will have an impact on the grid position, thats why I am proposing using an average of Q2 timing and fastest lap instead because during Q2 all cars are running low on fuel.

[Schummy]

Update after GP Spain.

TEAM    AUS MAL BAH SPA =  TOT
==============================
RAI-MAS 011 011 000 111 =  7-5
HAM-KOV 101 010 100 111 =  7-5
HEI-KUB 011 010 010 000 =  4-8
ALO-PIQ 111 111 101 111 = 11-1 <--
TRU-GLO 110 111 111 111 = 11-1 <--
ROS-NAK 111 111 111 010 = 10-2 <--
WEB-COU 000 101 111 111 =  8-4
BUT-BAR 000 111 100 011 =  6-6
VET-BOU 100 111 000 000 =  4-8
FIS-SUT 100 111 111 111 = 10-2 <--
SAT-DAV 111 100 000 011 =  6-6
==============================

• Alonso, Trulli, Rosberg and Fisi are squashing theirs team mates.
• Kubica and Bourdais are beating their somewhat supposed number 1 team mates. In Vettel's case some bad luck seems to be there, but we will be if it becomes significant.
• Extreme equality in Honda and Super Aguri.
• Slightly surprising equality in Ferrari and McLaren. Massa is fast and Hamilton looks a bit less impressive than last year.

[Sev]

I'd definitely agree with all of those points except the last:

I don't think Hamilton (statistically speaking) is doing any worse. I think his performances are similar to last year, only this year he is expected to beat his team mate. However, this more says things about Heikki's quality than anything.

[Schummy]

Adding some calculation over the former table. Race classification is too correlated with grid and especially with fastest laps to do a significance analysis with it. But taking in account both grid and fastest laps it can be reasonably done.

The first column is the "score" in terms of grid and fastest lap (NOT race classification). The second column is the "significance" of the score to mean one of the drivers is doing better than his team mate.

RAI-MAS 4-4  0% 
HAM-KOV 5-3 28%
HEI-KUB 3-5 28%
ALO-PIQ 7-1 93%
TRU-GLO 8-0 99.2% <==
ROS-NAK 7-1 93%
WEB-COU 5-3 28%
BUT-BAR 4-4  0%
VET-BOU 3-5 28%
FIS-SUT 7-1 93%
SAT-DAV 4-4  0%

Significance above 99% is widely used as relevant in technical and scientific conclusions. Significance above 95% is also used as a weaker level of "conclusion".

Just now, after GP Spain, Trulli has "proved" he is doing better than his mate, beyond the (reasonable) level of randomness inherent to GPs. Alonso, Rosberg and Fisi are close to reach that conclusion as well. Rest of teams are far from this (yet).

At the end of the season (18 races and, thus, 36 "points") a driver has to more than double his team mate to significantly beat him. It's to say, driver A has to outscore driver B at least 25-11.

I know this looks a mess of numbers, but I think the significance percentage of the above table is interesting.

[Pro Racer]

it's funny besides the Farce duo the others that are close to the 95% mark have first year team mates. alot of the other teams are closer then it looks on paper especially Ferrari (Kimi seems to be killing Massa) and Red Bull (Mark seems to be pushing DC into retirement).

[Born Racer]

Good work Schummy.

That last table seems to partly justify why I've been impressed with Trulli this year, because Glock is no slouch.

[vinegar]

Originally Posted by Greenback

Fuel load strategy will have an impact on the grid position, thats why I am proposing using an average of Q2 timing and fastest lap instead because during Q2 all cars are running low on fuel.

But not every driver goes flat out in Q2, especially at the front of the grid, as all you have to do is make the top 10. Equally, Q3 is fuel affected, and fastest laps can be distorted by track position - if you have a comfortable 10 second cushion, you have no real reason to squeeze the last couple of tenths out, whereas your team mate back in fifth needs to do absolutely everything he can. Since they introduced fuelled qualifying there's no real way to tell who's the fastest, except to average a large number of subjective impressions. But, as far as it goes, I think the selection used here is reasonable, although it might be worth weighting the overall classification to score 2 rather than one to account for the other factors being fluid.

edit to say - I see you think the opposite. I'd be interested to know why

[Schummy]

Update after Turkey.

TEAM    AUS MAL BAH SPA TUR =  TOT   G+L  SIG                   almost     proven
=============================================                         \   /
RAI-MAS 011 011 000 111 010 =  8- 7  5-5   0%                          | |
HAM-KOV 101 010 100 111 011 =  9- 6  6-4  25%                          | |
HEI-KUB 011 010 010 000 000 =  4-11  3-7  66%   +----+---              | |
ALO-PIQ 111 111 101 111 111 = 14- 1  9-1  97.9% +----+----+----+----+--+-|
TRU-GLO 110 111 111 111 111 = 14- 1 10-0  99.8% +----+----+----+----+--+--
ROS-NAK 111 111 111 010 111 = 13- 2  9-1  97.9% +----+----+----+----+--+-|
WEB-COU 000 101 111 111 111 = 11- 4  7-3  66%   +----+---              | |
BUT-BAR 000 111 100 011 011 =  8- 7  5-5   0%                          | |
VET-BOU 100 111 000 000 111 =  7- 8  5-5   0%                          | |
FIS-SUT 100 111 111 111 000 = 10- 5  7-3  66%   +----+---              | |
=============================================

G+L is the score just using Grid and Fastest Lap. SIG is the "significance level", >95% means is "almost proven" that A is better than B, >99% means is "proven" that A is better than B. The two "vertical lines" signals those levels. ["proven" used in an usual statistical sense]

Not surprisingly, Trulli has beaten Glock already, and Alonso and Rosberg are almost there. Webber, Fisichella and, surprisingly, Kubica are in the way.

[Schummy]

Originally Posted by vinegar

edit to say - I see you think the opposite. I'd be interested to know why

I need two or three "independents" measures of driver performance in a GP (weekend). Fastest race lap is correlated with race final position, so I cannot use both to do a statistical test (the significance level). I choose fastest lap instead race position because the later is more "randomly" affected by events (punctures, other drivers' crashes, mechanical failures, etc).

[vinegar]

True enough, but given that the circumstances of the race/fuel load/etc vary for each car in every race, and the ultimate aim of the weekend is to finish as high up as possible, by selecting fastest race lap instead of finishing position you're removing the only truly significant statistic from your comparison. perhaps an average of a driver's 15-20 fastest laps might be sensible, but even so it would be weighted against a driver who ran a one-stop race, for example.

Tricky business, isn't it?

[awrb]

I think the problem with using Qualy is shown by BMW. Kubica has not out-performed Heidfeld by that much this season, Nick is doing very well on points and has looked some what faster in the races. However, he is loosing points in this system because he always seems to go a few laps longer than Kubica in the race, therefore sacrificing grid positions for race positions.

I think this is why it appears Kubica it beating Heidfeld, when actually they just have a different appproach to a weekend.

[Schummy]

Originally Posted by awrb

I think the problem with using Qualy is shown by BMW. Kubica has not out-performed Heidfeld by that much this season, Nick is doing very well on points and has looked some what faster in the races. However, he is loosing points in this system because he always seems to go a few laps longer than Kubica in the race, therefore sacrificing grid positions for race positions.

I think this is why it appears Kubica it beating Heidfeld, when actually they just have a different appproach to a weekend.

I don't think so. Looking into Q2 data, Kubica beats Heidfeld 4-1, so Robert is being really faster than NH in Saturdays.

But the strategy bias in qualifying positions is a valid point. In fact I'm doing a stat with Q2 along the season, and I'm thinking about using it instead grid position. But, of course it is more work... , so I don't know when (or if) I'm going to change it.

[Schummy]

I have amended the table to change better grid position in Turkey for Fisichella instead of Sutil. Giancarlo was penalized because of a red light infringement, but this table is about performance. It changes some stats in the GF-AS row.

TEAM    AUS MAL BAH SPA TUR =  TOT   G+L  SIG                   almost     proven
=============================================                         \   /
RAI-MAS 011 011 000 111 010 =  8- 7  5-5   0%                          | |
HAM-KOV 101 010 100 111 011 =  9- 6  6-4  25%                          | |
HEI-KUB 011 010 010 000 000 =  4-11  3-7  66%   +----+---              | |
ALO-PIQ 111 111 101 111 111 = 14- 1  9-1  97.9% +----+----+----+----+--+-|
TRU-GLO 110 111 111 111 111 = 14- 1 10-0  99.8% +----+----+----+----+--+--
ROS-NAK 111 111 111 010 111 = 13- 2  9-1  97.9% +----+----+----+----+--+-|
WEB-COU 000 101 111 111 111 = 11- 4  7-3  66%   +----+---              | |
BUT-BAR 000 111 100 011 011 =  8- 7  5-5   0%                          | |
VET-BOU 100 111 000 000 111 =  7- 8  5-5   0%                          | |
FIS-SUT 100 111 111 111 100 = 11- 4  8-2  89%   +----+----+----+----   | |
=============================================