Bootstrapping F1 (2010) - Page 2

shiny side up! · 4 Aug 2010, 01:31 (Ref:2739146)

Looking forward to it!!!!

Alfa Fan · 4 Aug 2010, 01:35 (Ref:2739147)

Would using a linear regression model, or possibly a logit/probit style model be much more appropriate here?

Schummy · 4 Aug 2010, 02:37 (Ref:2739155)

Quote:

Originally Posted by Alfa Fan

Would using a linear regression model, or possibly a logit/probit style model be much more appropriate here?

Using what regression variables?

Alfa Fan · 6 Aug 2010, 01:11 (Ref:2740139)

Quote:

Originally Posted by Schummy

Using what regression variables?

You would have probability of a win as the dependent variable and then regress it against any number of influencing factors using a logit/probit/similar model.

The dependent variable would be something like the probability of a win in any given race or the expected number of wins in a season (statistically identical) and then regressed upon any number of available factors....like driver/teams historical performance, performance to date, weather adjustments etc. Would be fascinating to see the results, although I'm not sure what data you have available?

Then for probability of the title, this would be a bit more complicated. But essentially you use historical data as well relevant current information (like tracks, races known to favour certain drivers/cars, a rolling average of recent form or whatever). In fact thinking about this has got me interested enough to hopefully go and collect the data at some point!

Schummy · 6 Aug 2010, 02:04 (Ref:2740158)

For title probability I'm a bit skeptic of the regression approach even accounting for the "S"-shape of probability by means of a logit (or probit) scheme.

You have to collect variables more or less related and then you have to efficiently prune them, a posteriori, to avoid overparametrization.

BUT... it could have some advantages. It count be mounted like a inferential method, giving error estimations. Most of all, the interesting thing I foresee is finding some interesting relations between title and some variables. Some surprises could be found.

I hope you do it and put the results, so we, the numerical mad people here, can have more fun. No matter if grand results are achieved or not, the results would be interesting to read.

Let's go, Alfy! Go for it!

Chatters · 6 Aug 2010, 04:48 (Ref:2740196)

Speak English you two.

Scrut · 6 Aug 2010, 04:54 (Ref:2740198)

Quote:

Originally Posted by Chatters

Speak English you two.

Totally agree, Chatters.

Schummy · 6 Aug 2010, 12:12 (Ref:2740345)

Skropaczenski awaanang werdkloszer oy aanaklappach ewuk zyllaq shotto bamks!

To be perfectly clear.

phoenix · 6 Aug 2010, 13:07 (Ref:2740372)

OK - so sometimes this thread is like listening to terminology that they use in Star Trek - but I like the graphs

When's the next graph coming Schummy?

**Born Racer** · 9 Aug 2010, 12:05 (Ref:2741851)

Teflonso reckons his chances of winning the championship are about 50%, which is a lot higher than the percentage forecasted earlier in the thread. http://www.autosport.com/news/report.php/id/85905 Although maybe it's about as accurate a measurement as the six-tenths he brings to a team.

Schummy · 9 Aug 2010, 13:26 (Ref:2741892)

Between Alonso and Paul Octopus I prefer Paul as forecaster

bravo · 1 Sep 2010, 10:17 (Ref:2753102)

Shummy!! We need new graphs, especially with the Nil points for Vettel and Button at Spa.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
F1 2010 (Codemasters)	ECW Dan Selby	Virtual Racers	630	24 Jun 2011 10:38
[Official] 2010 F1 Predictions and Fantasy F1 Competitions	Ralf's Girl	Formula One	4	8 Mar 2010 17:10
Bootstrapping F1 (stats)	Schummy	Formula One	39	2 Jun 2007 13:49