[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums

Improved Cube Handling in Races

Posted By: Timothy Chow
Date: Tuesday, 24 June 2014, at 2:02 a.m.

In Response To: Improved Cube Handling in Races (Axel Reichert)

Axel wrote:

27 billion possibilities doesn't sound like that many to me.

It is. I would guess that the genetic algorithm I used to sample this design space covered maybe 100000 of them at most. Now if I was to extend my search to the complete design space (i.e., by a factor of 270000), I have to reduce the database of position by the same factor, with only about a quarter of a position left ...

You're right that the chances of missing the optimal solution are minuscule, so my point is academic.

However, I believe that my point is still technically valid, so let me restate it. Since most strategies are going to be garbage, it should be easy to show that they are garbage with only a small random sample from the database. You don't need to check anywhere near the whole database to get a quantifiable probability that a particular strategy is complete garbage; that's the beauty of random sampling. Furthermore, you don't need to run anything complicated like a GA. I'd have to think about this a bit to be sure, but I think you could do this kind of exhaust on a GPU. Even if not, I'd think that 10 microseconds should be plenty to test any single strategy on a small random sample of the database, and rule it out with 99.99% confidence.

Messages In This Thread

 

Post Response
Your Name:
Your E-Mail Address:
Subject:
Message:

If necessary, enter your password below:

Password:
 

 

[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums is maintained by Stick with WebBBS 5.12.