Author: Dann Corbit
Date: 08:07:10 11/12/98
Go up one level in this thread
On November 12, 1998 at 03:12:35, Itay wrote: >I am student for math and computer sciences in T.A University Israel. >I am also a Chess fan trying now to create my own chess program. Right >now I finished the first step of implementing all the chess rules on the >program, and started to operate it. >The program finds the first possible move. But I bump into problems when >I try to add algorithms of Artificial Intelligence. Look at the Alpha Beta routine. It is a form of AI, and very nice. Most other AI stuff is not very functional right now. >I ask for your help and your advice . Go to "computer chess resource center" "computer chess links" "programming methods and studies" There are a bunch of links to AI chess programming studies there. You can even find source code if you follow all the links. I have source code for several versions on my machine. Prepare to be disappointed. These AI programs are really weak. You will find that it is surprisingly easy to beat them. Just wait for a horrible blunder and pounce. I think they are going about it all the wrong way (of course, they are trying a 'pure' approach, and I just want to make something stronger). Contemporary programs often learn by tagging a move that cost it to lose as "don't ever play this move." I think that is a bad solution also. Really, you should just update the won-loss statistics for that move. In that way, you might still find interesting solutions. It could also be valuable to post-process at very long time controls to find out what really went wrong and update your database of positions with the new information. It may lose a lot more chess games initially, but eventually, that method would become stronger IMO. I am going to write an AI chess program of sorts. But not the traditional kind. The knowlege will be captured in a database. At any rate, I would suggest that you try the link I mention, and read all of the papers. You will need a postscript printer or GS-View or something of that nature, because most of the publications are in postscript format.
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.