4 replies to this topic

[photon]

Anyone here ever considered trying to write algorithms for predicting race strategy?  Was just curious if there were any CS, Machine Learning, math etc people around here that have fiddled around with such a thing.  For example, predicting what tire strategy is best and when to pit etc.

[Jovanotti]

To add to your questions (if I may ), I also wondered a few days ago if there is any racing related software available on the net (can be anything really, lap time analysis, strategy calculator, calculating where there are gaps opening in the field,...)?

Edited by Jovanotti, 14 May 2014 - 14:45.

[Greg Locock]

On a slightly related note, some solar car teams used to obsess about plotting other cars position relative to their own. Since the optimum strategy for a solar car is entirely independent of the behavior of other cars that never struck me as a worthwhile endeavor, except that it told you about sun and wind up and down the road.

 

However an F1 race is not so easy, I wonder if it is worth putting any real complexity in, or just taking the current state, using a few rules of thumb about tire degradation and fuel load, and then reset the model at the next safety car.

[Lee Nicolle]

You dont need any fancy math! Too much computer crap already, often  to the detriment of the team. Just commonsense and watch the pace car and your main opposition. 

Look at many succesfull Nascar crew chiefs, some probably could not sat algorithim,, but they can get a car too the flag first !

[Joe Bosworth]

I have written such a programme.  It is actually quite simple to write a spreadsheet programme for your own car, assuming of course that you have decent input data.  The old GIGO theory holds here in spades.

 

Just as Lee says above, you doin't need fancy math. Just keep track of forecasted lap times based on trye degradation and fuel lightening including pit stop fixed and variable times based on fuel load and tyre change times. If you have good data for the affects of track temperature you can plug your best guess of weather conditions over the race time into the spreadsheet as well.

 

Keep a running sum of  forecated race time at the end of each lap and keep an eye on lowest total race time.  Be ready to plug in real lap time deltas for yellow flags etc. if you are going to use at the track side.

 

It gets a bit problematic when you start to plug in your position aand strategy versus say two competitors as you have to then play with the problems that you don't know their degradation and pitsyop data so you have to get some good spy data and/or use a good overlay programme to your spreadsheet.  I use and can recommend @Risk which lets the spreadsheet compute most likely scenarios, especially for passing other cars and time lost/taken in passing.  This is a variable almost impossible to accurately forecast before the raace so you ignore it in setting your base strategy.

 

I have never tred to go past two competitors just because the data input gets to time consuming for one guy and I have never been in a position where one could get two or three other inputs to plug build multiple scenarios beyond you plus two others.

 

If you are spreadsheet literate which includes developing sheets with live inputs it won't take more than an hour or two to develope your home team scenario.  It also helps to be statistics literate if you have lots of historic data to crunch to get better inputs.  Remember GIGO!!!