Author: Robert Hyatt
Date: 20:08:44 05/28/03
Go up one level in this thread
On May 28, 2003 at 16:39:29, Dieter Buerssner wrote: >On May 28, 2003 at 03:36:10, Ed Schröder wrote: > >>http://www.rebel.nl/db2.htm > >I think, it is disputable, that this shows the longest line. Actually, I believe >the question for the longest line, to see the draw is practically not answerable >(is the an English word?) in a position like this. There are many different >lines, that have equal or close to equal material balance. one could cut a line >with an argument like "Now black has easy play for draw", or many similar >arguments. How to define "easy", is not so clear, however. In parts of the >lines, white will have a clear advantage (in material) and everybody would >agree, that the only chance for black will be perpetual check. In those lines, >we can find the longest line to draw. Again, we would probably find lines, where >"clear advantage" [for white] is not that clear anymore, and could discuss for a >long time about this ... > >Regards, >Dieter I think the thing that is "missing" is the ability to statically recognize "this is obviously a perpetual". IE I remember looking at a position once where the program ran into the king trapped on the g/h files, with a rook checking it continually. You might think it was an easy to find draw, but when the king went from g1 to h2 to g3 to h4 to g5 to h6 to g7 to h8 to g8 to h7 to g6 to h5 to g4 to h3 to g2 to h1 that is a _long_ set of extensions. And they all occurred at the end of a long variation (I think this might have happened in a Crafty vs Yace game on ICC a few years back in fact). As a human I instantly saw that this was going to be a perpetual, and I didn't understand why Crafty couldn't see it as I was thinking of kg1 kh1 kg1 and there you go. But that diagonal up and back crap pushed the first repetition a _long_ way out... Having to search that subtree to conclude "draw" is simply too costly. I have also seen the stalemate trick where the rook checks the king forever and the king can't take because the other king would be stalemated. And those can cover the entire board before the first position is repeated, going _way_ beyond 60 plies, meaning the 50-move rule has to terminate the search at 100 plies. Not going to happen.
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.