Probability of win (warning: strange calculations!)

Schummy · 9 May 2005, 14:24 (Ref:1296683)

I have done (strange) calculations to try to predict the probability for win in the next (generic) GP.

I take in account past wins (always in this season), 2nd positions, a discount factor to give more weigth to recent GPs and less to older GPs. I also have assumed that at the beginning of the season every driver had the same a priori probability of win (5%).

After GP Spain and with those calculations I arrive to this prob table for driver's win and team's win:

Code:

Drivers:
Alonso      37%
Raikonnen   18%
Fisichella   9%
Trulli       6%
M. Schum.    5%
Barrichello  3%
Any other   22%

Teams:
Renault   46%
McLaren   20%
Ferrari    8%
Toyota     8%
Any other 18%

Table of results (1st=1 , 2nd=0.25) after the 5 first GPs:

Code:

Importance   0.41 0.51 0.64 0.8  1
-------------------------------------
        FA   0    1    1    1    0.25
        KR   0    0    0    0    1
        GF   1    0    0    0    0
        JT   0    0.25 0.25 0    0
        MS   0    0    0    0.25 0
        RB   0.25 0    0    0    0
-------------------------------------

After the next GP, those numbers will have to be updated with the new info.

I DON'T claim it to be particularly accurate

It is just done for "entertainment"

(yes, I'm a freak)

All is needed to do those calculations are a calculator, paper, pen and a rainy afternoon

Dani Filth · 9 May 2005, 14:51 (Ref:1296708)

what can i say . typical Schummy

Kicking-back · 9 May 2005, 15:05 (Ref:1296714)

There are so many variables in racing it can never be a statistical exercise - but interesting anyway

Schummy · 9 May 2005, 16:23 (Ref:1296785)

Just au contrairie, dear K-b

, statistic comes when there are too many variables to do a deterministic calculation!

But, anyway, you, Dani and me know this kind of number crunching is just to entertain ourselves

Don K · 9 May 2005, 16:33 (Ref:1296789)

Did you do any check whether this kind of calculation would have given reasonable results in the past (for instance in 2004 or 2003, after about 5 races)?

It would be very nice to have some calculation as simple as this (with very few input data) with a decent prediction value ...

dcp2685 · 9 May 2005, 16:47 (Ref:1296813)

Yes of course. At any given race it's 50% reliable. Either the predicted driver wins, or he doesn't.

Bah statistics.

supermagnet · 9 May 2005, 18:04 (Ref:1296898)

Quote:

Originally Posted by Don K

Did you do any check whether this kind of calculation would have given reasonable results in the past (for instance in 2004 or 2003, after about 5 races)?

It would have been spot on last year

Inigo Montoya · 9 May 2005, 20:12 (Ref:1297008)

Are you using these stats to help you in the predictions competition Schummy?

Hugh Jarce · 9 May 2005, 20:17 (Ref:1297013)

I must be getting old - I'm not only struggling to 'get it' - I'm blowed if I can even read the small print.

Does one of the words on those tables say 'impotence'?

Schummy · 9 May 2005, 21:20 (Ref:1297059)

Don K: I have not done any fine tuning with past season data (lot of work!). I think I'm going to try to check it along this season. But to do it I need more rainy days and we are approaching summer...

Inigo: Sadly I don't play in the F1 game

. I run two games in other forums and I don't have more time! Anyway, I don't think these numbers can help me in guessing games (c:

Hugh: Yes, it reads "impotence", it is a sort of Rochard test

Logrence · 11 May 2005, 13:48 (Ref:1298329)

I think you need to work out another way of working out the rest of the drivers' chances of winning the race - statistically it may be 22% likely that any of the other drivers might win the race, but in reality I'm thinking snowballs, hell, that sort of thing.

Schummy · 11 May 2005, 23:28 (Ref:1298768)

22% is the cummulative prob of wins for Webber or Heidfeld or Ralf or Montoya, etc. So it is not so unlikely as it appears.

But I understand your point, our perception of probabilities always difer somewhat of whatever calculations we do. We try subjectively to add colateral info about tracks, drivers teams, tyres etc... Some of them are right, some of them wrong...

Don K · 12 May 2005, 14:20 (Ref:1299206)

We also tend to misunderstand percentages.

For instance, many people would think that the chance of the Football World Cup to be won by either Brasil, France, Germany or Italy would be greater than 99%.

Historically, the chances of the Footbal World Cup to be won by either Brasil, Germany, Italy, Argentina, England or France seems to be only 88% ...

---

On the other hand, historically there seems to be:
- 35% chance that the World Cup is won by the organizing country
- 29% chance that the World Cup is won by Brasil
- 94% chance that the World Cup is won by either a team from the continent in which it is held, or by Brasil

So a simple calculation can give us some "realistic" chance in many cases.

Logrence · 13 May 2005, 13:17 (Ref:1299890)

It's not a cumulative probabability though, is it - it merely says "any others". That means I have a 22% chance of winning Monaco if I enter in one of the Minardis...

Schummy · 22 May 2005, 18:41 (Ref:1307374)

Updating these "strange" probability-of-win calculations after Monaco, we have for the next GP:

Drivers:
KR 29%
FA 28%
GF 6%
NH 5%
JT 5%
MS 4%
RB 3%
Others 20%

Teams:
Renault 34%
McLaren 32%
Williams 7%
Ferrari 7%
Toyota 7%
Others 14%

Kimi has got Alonso as the most likely winner, and Mac is now almost equal to Renault. Slight pass forward for Williams and Heidfeld, respecting others teams.

Nothing terribly surprising, lol

Schummy · 31 May 2005, 12:07 (Ref:1315665)

Updating after Nurburgring:

Probability of win:

Drivers.
FA 36%
KR 22%
NH 8%
GF 5%
JT 4%
MS 3%
RB 2%
Others 20%

Teams:
Renault 41%
McLaren 23%
Williams 9%
Toyota 5%
Ferrari 5%
Other 17%

Now Heidfeld is the third more likely winner and Williams is advancing from the rest of teams.

Skam85 · 1 Jun 2005, 09:01 (Ref:1316532)

Great stuff Schummy, but one question niggles at me-do you have a job?

Schummy · 1 Jun 2005, 10:57 (Ref:1316647)

My job is to serve all of you, masters

But really the question is how many rainy days I have had lately? lol

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
The Best 2005 GT1 Car? It is Corvette by my calculations….	David	Sportscar & GT Racing	30	19 Dec 2005 22:29
Very strange to see.......	steve nielsen	Formula One	10	10 Sep 2001 19:07