Author: James Robertson
Date: 11:31:22 10/24/98
Go up one level in this thread
On October 24, 1998 at 09:55:45, Robert Hyatt wrote: >On October 23, 1998 at 21:07:34, James Robertson wrote: > >>On October 23, 1998 at 20:40:03, Robert Hyatt wrote: >> >>>On October 23, 1998 at 00:45:33, James Robertson wrote: >>> >>>>On October 22, 1998 at 23:14:56, Robert Hyatt wrote: >>>> >>>>>On October 22, 1998 at 21:45:54, jonathan Baxter wrote: >>>>> >>>>>>How do chess programs detect perpetual check? Do they just extend on sequences >>>>>>of checks or do they have specific features to detect likely perpetual >>>>>>situations? >>>>>> >>>>>>Jon >>>>> >>>>> >>>>>I don't do anything special other than checking for repetitions in the usual >>>>>way. I extend on checks, and again if there is only one legal move to get >>>>>out of check. Other than that, nothing special in my code at all... >>>> >>>>Wasn't Deep Blue unable to detect perpetual check in the second game of it's >>>>Kasparov return match? And Kasparov then made the foolish mistake of believing >>>>the computer wouldn't miss something like that? >>>> >>>>James >>> >>> >>>Kasparov didn't see it either. It was a full 60 plies of search to see >>>the repetition... >> >>I suppose it is obvious he didn't see it; he didn't play it. :) >>Many perpetual check positions require many many plies to see them to their >>conclusion (draw), and yet humans spot them all the time..... >> >>James > > >this is a problem. And we solve it differently. IE I don't search for >a repeated position, I search for a way *out* of the checks. And if I >don't find it, I call this a draw. Crafty has to do it the other way, and >sometimes the draw is so deep it wouldn't see it in a year, while a human >would see it instantly. IE a position where white's queen is at a1, a >white pawn at a2, and the king + black king/queen anywhere on the board >where the black king isn't already in check... the black queen will >run the white king all over the board before the position is repeated. >But a human would look and say "no way to build a place to hide here, so >there is no way to escape the queen checks... this is a draw" > >Computers just don't handle this very well... Yes....... In the Deep Blue - Kasparov position, the board was cluttered enough it was difficult to realize there was no way for the White King to escape the checks....... I guess this one is a combination of both methods (hiding and repetition), and therefore neither player saw it. :) James
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.