Author: Robert Hyatt
Date: 05:17:42 03/22/99
Go up one level in this thread
On March 21, 1999 at 18:52:18, Mark Young wrote: >On March 21, 1999 at 18:32:10, Baldomero Garcia, Jr. wrote: > >>Well, I'd say forget about the endgame tablebases. Really, in how many endgames >>will it make that big >>a difference? >>A much more significant aspect would be if a programmer would be able to teach a >>computer program >>to play endgames well. >>IMHO, adding more opening moves or more tablebases doesn't necessarily make the >>program stronger. >>It has to "think" less moves. > >This is a different issue, but I agree with 5 or less man EGTB, the advantage is >pretty small. > >I think it would be a total different story if a program would have a full 9 man >EGTB. Playing well is one thing, playing perfect from 9 man on down is another. the point will be that programs that probe _in_ the search will do even better, as they don't wait until there are just 9 pieces to probe. They search positions with 16 pieces and then trade into won 9 piece endings. When we get to 9, the computer will be _very_ difficult to beat (very difficult is pretty close to impossible). IE as in the "chinook" program in checkers, where from the opening position (a couple of years ago) they were hitting about 70% of the positions ending up in their endgame databases. They don't lack a lot to make that 100% and then they have "solved" checkers. While we won't do this often is chess, the result might be the same eventually.
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.