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

BGonline.org Forums

roll out setting

Posted By: Matt Cohn-Geier
Date: Friday, 22 August 2008, at 10:08 p.m.

In Response To: roll out setting (Stick)

On the whole, you roll something out w/Snowie, you roll something out with GNU, you get the same results.

I'd agree with that. I don't think that this is in conflict with GNU being a better choice than Snowie in some situations and Snowie being a better choice than GNU in others. In general I am more inclined to trust GNU in races, match play, etc., and I am more inclined to trust Snowie in back games, etc. Of course in 99% of situations you will get the same result with either bot.

E.g., we pretend that the results of rollouts are gospel when we agree with them, and when we disagree we say we don't trust the rollouts and go with our intuition. Then we're told that gnubg is better at some positions and Snowie at others, etc., etc.

This is clearly a bias but it's not without any kind of foundation. There are rollouts and other data to back up the fact that Snowie is typically better at back games with many checkers back than GNU. If you would like, you could in theory collect similar data on any position, but in practice, you're limited by time constraints.

And are there any circumstances where a 2-ply evaluation is better than a 3-ply evaluation? If so, why?

I am conducting some rollouts on this. 2-ply cube appears to be slightly better than 3-ply cube from the look of things. 3-ply checker appears better than 2-ply checker, but I am not sure by how much.

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.