Author: Bruce Moreland
Date: 10:24:55 02/04/03
Go up one level in this thread
On February 04, 2003 at 09:08:50, Mike Carter wrote: >Dann - thanks for your reply. I've seen most of these articles before and I'll >read them again. What I was hoping to get was a one-paragraph explanation of >how to implement the alpha/beta concept... specifically, when/how often does the >evaluate routine get called? I saw a program that, for 4 ply, was analyzing >8K-10K total nodes. Logically this seems impossible to me (to be able to prune >the move tree that much). I'll sift through the websites you sent, and >hopefully I'll stumble across something that turns on the light bulb for me. >Thanks again Dann! Min-max provides you with a bf of about 35. Alpha-beta gets this down to about 6. Null-move forward pruning reduces this further. If you use a quiescent search, if you are ruthless about pruning what appear to be bad capturing lines, you can reduce total nodes by a factor of maybe 2, perhaps more. You haven't got the basic thing working yet. You have to get that running first. The basic thing will play like poo, which is perfectly okay. Once you get that working, you can spend more time trying to make it play well. In the mean time, if you have something, anything, that will play legal chess, hook it up to Winboard and get it playing on ICC. Weak programs are extremely popular, and I encourage everyone who has one to get theirs playing online. Last time I wrote a chess program, I had it online playing before it had any eval more than rudimentary piece-square. This was the right way to go, it was a major blast to see the thing being silly so soon after inception. bruce
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.