Season evolution for drivers and teams

[Schummy]

I have done a bit of calculation about who has the momentum and how drivers/teams have gone thru the season so far.

I know all this is a bit "heavy" and possibly boring , but if you are interested in "numbers" go ahead!

(Disclaimer: the author is not responsible of any lesion or malfuction of the brain of the reader as consecuence of reading of this post)

I order to do it, I have done a moving average (exponential, smoothness parameter chosen especifically) with point data of the first seven drivers and the three top teams.

The resulting graphs are too large to be allowed here (although are a small gif file), so I have to put here just the link:

http://summer.phpwebhosting.com/~sta...Statistics2003

The teams graph is nice, showing sucesive Mac, Ferrari and Will domination thru the season. Also it shows as none of the team has really dominate in any part of the season (except for short time).

In drivers graph one can see the early Kimi domination, and Michael bad start, then Michael's resurgence and Kimi's (it's to say MP4-17) gradual decline, and, finally, Williams pair coming to the top. Also one can be Rubens has been quite regular thru the season, and Alonso has been pretty regular too, but with a bit of decline in results.

Just for the sake of fun, I have done a prediction for final positions in this year WC. Current forecast average points for the next GPs (not taking in count it is Hungaroring in particular) is as follows (per GP):

JPM 7.2
MS 5.4
RS 4.3
RB 4.1
KR 3.9
DC 3.8
FA 3.2

JPM has the best momentum, but MS is still second. Luck, as always, has taken part in all this (KR, RS and RB's strat line accident in Hock, DC's good luck just for it, Michael's good/bad luck with that accident and then the tyre...), but that's how points have happened.

Adding those "predicted" points per GP for the last 4 GPs and adding all it to the current points for each driver, we get a posible WC final classification:

JPM 93.7
MS 92.6
KR 77.4
RS 70.0
RB 65.3
FA 56.7
DC 56.2

JPM and MS are heading (if current momentum of form goes on) to one close final classification for WDC. KR, RS and RB would have to resign to fight for third final position, and FA and DC will fight(?) for sixth.

About teams, current points per GP "predictions" are:

Williams 11.4
Ferrari 9.5
McLaren 7.7

And "predicted" final posititions in WC are:

Williams 163.7
Ferrari 157.9
McLaren 133.6

Most of people thinks Williams will get the team championship, but I abide to these data, and I think it will be close with Ferrari.

It is so close in drivers and teams championship that very probably luck will decide who is the winner: weather, track temperature, tyre choice by Michelin/Bridgestobe, mechanical troubles, incidents/accidents,...

[avsfan733]

you are so muc hsmarter than I am that it is not even funny. I can't quantify F1 past, he won he finished second... That is just amazing and after studying it and figuring it out, its incredibly interesting to look at it that way, even with the math though to predict it it is still so close.

[Irv the Swerve]

So, for example, if I were, hypothetically speaking, going down to the bookies in about 20 minutes, you would say that mathematically and statistically based on recent trends, Montoya would win the WDC.

Thats good enough for me.

Do you work for a bookies Schummy?

[Schummy]

I'd say JPM and MS are almost equal in terms of probability. BTW, neither of them has the guarantee of being crowned as WDC, they just are the most probable champions.

But all that is according to the model used here. It is relatively accurate but can be improved taking in account not the points themselves but the "pace" in the GP (maybe fast laps, quals, etc). This way luck role would be lesser.

For example, if Schumacher's tyre had not deflated, he perhaps had finished in third position. If it had happened, the prediction offered in this thread had put MS as the WDC favourite. It is so close as for results be affected by any small circunstance. Of course other model can be used appart from this kind of moving average (you know, the more "technical" is the approach, the less concrete is the answer )

My personal opinion is it is very close (as we view it today August 14), and it will be surely decided by circunstances, not by pure pace or form. In this frame of things I'd bet for MS, as he is more experienced and possibly is a bit better prepared if circunstances affect him. But if a bolt get dettached, a tyre degrades or something like that happens, nothing can he do.

By a tiny margin my favourite is MS, but I would not put a lot of money on it! Anyway, betting a less probable outcome is also atractive because is is rewarded with a greater amount of money (c:

[If on Sunday 17 MS is well back in the grid and JPM is in the first row, I'll change my "vote" even before the race has started]

[Mr V]

Schummy, you are far too clever for me, however, any list with JPM's name at the top is fine in my book

[Peter Mallett]

Originally posted by mr v
Schummy, you are far too clever for me, however, any list with JPM's name at the top is fine in my book

You are so sad

[Bononi]

That's mr v...

[Logrence]

That's the lot of you...

[Mr V]

Originally posted by Peter Mallett
You are so sad

Well, we all have our crosses to bear

[Schummy]

(LOL at all that cross fire! )

At least, miraculously I'm being kept out of the shooting notwithstanding my "dangerous" nick

[JohnSSC]

Schummy, I am in awe! I have seen less detail in a proposed takeover bid of another business!

Tell me (seriously) do you try to come up with a standard deviation to try to take into account the odd tire deflation or the random wingnut running onto the track (see Silverstone) or do you try to keep the calculation simple and let the overall finishing positions (which have already been impacted by the random events) dictate the forecast?

For example, in working with sampling routines to perform a quality analysis, we would ask for a standard deviation so that we knew within plus or minus so many points what the accuracy rate would be and the statisitical certainty that it was reliable (93% or 95% for example).

Don't let anyone give you grief regarding your nickname!

PS- you've done this sort of thing before, haven't you? .

[R]

Some great work there, Schummy! The best price for TGF seems to be 1.60, while Montoya can be found at 3.50, so if you're betting, Montoya looks like the better value bet, given your estimations of how close it'll be.

(Bloody hell, mr v, I liked your old username better. With caps, you look so... big!)

[neilap]

Schummy, good work. Thanks!!

[ralf fan]

Good work Schummy!!thanks for the stats

[Jordi]

Man, what about some stats for the lower positions? Will Jacques score points?

[R]

Do birds fly? Do bees hum? Yes, of course he will.

[Schummy]

Thanks people for kind words, even for naughty words
When I do something with numbers that I think can be somehow interesting for F1 people here, I try to post it. Sometimes it is a bit dry but I always try to warn people at the beginning of the post!

-------------------------------
Jordi:

Instead JVill I can do something about Oriol Serviá, I think you are a supporter of him (from other forums here )

BTW, days ago I heard Oriol on Eurosport doing some the commentary of a CART race, possibly you did it as well.

---------------------------
JohnSSC:

Thank you! I try to keep those calculations rather simple, or they will suck me as a big black hole!

I other ocassions (different F1 things) I calculated deviations in order to do a estimation of probabilities or confidence intervals (usually thru a gaussian model), but I avoided it this time to put a "fast" stat, more than a detailed (and time-devouring) one.

Anyway, I've used a "smoothness coeficient" that I had gotten earlier doing calculations for the last 3 or 4 seasons about what coeficient could forecast better. So, it had more earlier work done already.

I also have done a WDC final predictions based in an (relatively) easy but statistically sound technique called "bootstrap". But possibly I don't put it in the forum because I don't want to "overdo" the calculation thing here or people will begin to think I'm Prof. Tournesol (Calculus) If I can make a "nice" graph out of it, I'll put a new thread with it.

[JohnSSC]

Wowser!

So Schummy, just for fun have you ever taken a season with a known result (say 2001) and then ran your calculations? Do you come up with the same result, or is it "off" by the standard deviation you put in for "smoothness."

This is cool stuff - I wish I were better at the actual application.

. Oh gosh! This was my 1,000th post! Who would of thought it would be on statistical probability modeling???

Cheers everyone!

.

[Schummy]

JohnSSC: Congratulations for the 1000th post!

I have used some past seasons (i.e. 2001) and then I've forecasted every GP of every "mayor" driver. I have done the usual mean of squared deviations for all that, for each value of coef of smoothness (well, for "many" values).

Doing it, I arrive to the "optimal" coef to forecast a driver GP taking in account past GPs of that season. Optimal forecast for the next GP is, in fact, a measure of how good is currently that driver/car. It is what I wrote as "predicted points per GP".

The value of the optimal mean of squared deviations is an estimation of the variance of predictions, but it is better to calculate a relative variance that take in account the size of probable points for a given driver.

All this I do for "fun" and it can be time consuming. But when I get some interesting results/graphs I get rewarded. It is like avery hobby: absurd and time consuming but rewarding

[Maxmil]

Very interesting Schummy. Two things strike me: (1) If you can get Montoya at 3.5 to 1 by all means take some, and, (2) the analysis sure points out the disaster at McLaren this season. DC is in the basement all the way, and Kimi has faded like the morning dew in a drought. The remaining question: Is it the car or the drivers?

[JohnSSC]

Being able to "see" the performances charted out certainly provides a starting point for the conversation of "what went wrong" or "what made it better" - still, great stuff.

Like every hobby though, "If you can't hear the music, you won't understand the dance!"

[R]

Maxmil, in fractional odds it's actually 2.5 to 1, or 5/2 (if you bet 2 units, you will win 7 units if he wins = profit of 5 units). In decimal odds, that's 3.5, because the stake is always included in decimal odds. Likewise, TGF is at 1.6, which in fractional odds means 3/5.

For McLaren, I think it's the car, and not the drivers. They're behind on development, and the never ending MP 4/18 saga, and the uncertainty of whether or not it will ever actually be used in a race, can only have negative effects on the drivers IMO, no matter how much they're trying to not let this affect them. That was supposed to be this season's car, for crying out loud! (Not to you, but to McLaren )

Schummy, have you considered taking into account previous performances on the same track? Some drivers are better at certain tracks, and not so good at others, for example JPM has been - pardon the expression - pish poor at the Hungaroring in his past two appearances there.

[Schummy]

R:
It is a interesting factor. I remember I used to think that somehow Patrese went well in Imola; Max Biaggi is strong in Brno (ummmm... let's see this year...); MS seems at home in Spa/Suzuka... a lot of examples.

But, again, it involves pretty much "forix-ing" and comparing past seasons perfomances with different cars/tyres, etc, it is not easy to do in a standard way, but possibly it can be cracked in some non-standard way. What I have in project is to do an analisys for a particular driver and a particular track... perhaps in a long rainy autumn evening?

Having just 4 races, this effect is worthy, but if we were doing predictions more in a long term (in middle season or so) it would not so important because different circuits would tend to level one with another.

JPM looks a bogus track for JPM, Suzuka seems to be good for MS, Monza can favour BMW power (but also can favour Ferrari aero low drag) someone can say what track (of the remaining four) favour JPM or MS?

[Inigo Montoya]

Careful Schummy, or people will think you have an unfair advantage in the predictions competition Thanks for the interesting stats!

[Smokey 6 litre]

thats a good set 'o' graphs, good idea about the disclaimer as well, had a slight seizure while reading it.