Author: Robert Hyatt
Date: 10:15:31 06/17/03
Go up one level in this thread
On June 16, 2003 at 23:46:15, Keith Evans wrote: >On June 16, 2003 at 23:23:41, Robert Hyatt wrote: > >>On June 16, 2003 at 02:50:49, Ricardo Gibert wrote: >> >>>On June 14, 2003 at 18:00:30, Robert Hyatt wrote: >>> >>>>On June 13, 2003 at 12:03:58, Michael Vox wrote: >>>> >>>>>http://www.clubkasparov.ru/521772350.html?462691585533321 >>>>> >>>>>One could argue chess endgame tablebases play the endgame like god, but not this >>>>>article.... >>>>> >>>>>Enjoy :) >>>> >>>> >>>>The author is an idiot. >>>> >>>>a 5 piece endgame _counts_ the two kings. He is not counting them. >>>> >>>>He really thinks he is probing what we would call a 7 piece ending, which >>>>is _years_ away from reality. >>> >>> >>>At no point in the article does he ever do as you allege. He always counts the >>>pieces correctly. >>> >>>We all make mistakes, but I don't think we should therefore brand all of >>>ourselves "idiots". Do you? He is a GM after all, so don't you think you calling >>>him an "idiot" a little extreme? >> >> >>Perhaps "computer chess idiot" would have been better? >> >>His entire article is based on incorrect information. >> >>A 5 piece position is _always_ played perfectly by a program. But when there >>are more than 5 pieces on the board, perfection goes away even when probing >>5 piece tables after captures. >> >>In his text, I get the impression he is saying position two should be played >>perfectly. Yet it has _seven_ pieces on the board. Tables work miracles, >>but they don't make the impossible possible, yet... > >Nevertheless for position 1, after 1.Bd1 Kg8 2.h7+ Kxh7 3.h6 Kg8 4.h7+ Kxh7 5.h5 >Kg8 6.h6 Kh8 7.h7 Kxh7 there are only _five_ chessmen on the board. So if he has >tablebases enabled, then what _should_ the engines return? I don't have 5-men >tablebases available, so I don't know. Is his analysis incorrect, or is he >pointing out a bug or setup problem with Junior and Fritz? The problem is this: If the position _starts_ off with 5 pieces, it will play _perfectly_. If it starts off with more, it might not. IE it might find a forced drawing line because of following a series of checks to a drawn 5-piece ending. But if it didn't take some of the checks, there is a deeper line to a won ending. If the program starts off thinking it is behind, it will take the draw because it can't search deeply enough to reach the won position. If it thinks it is ahead to start with, then it will avoid the draw and perhaps reach the won position although that is not certain since it is deep and it might make a mistake still and draw. The moral is that if there are moves _prior_ to hitting the EGTB, then there are no guarantees that it will play perfectly at all... > >Also why do you get the impression that he implied that position two should be >played perfectly? He does wonder this about position six which does have 5 >chessmen. And what is the answer for position six? That it _will_ be played perfectly. When it wasn't, that clearly shows that there is something wrong in the way the test is done, because a 5 piece ending can _not_ be played incorrectly with 5 piece tables available. It is simply impossible.
This page took 0.02 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.