Author: piet de hoop
Date: 12:24:12 11/19/01
Go up one level in this thread
On November 18, 2001 at 18:21:34, Dieter Buerssner wrote: >On November 18, 2001 at 16:48:08, piet de hoop wrote: > >>On November 18, 2001 at 15:54:23, Marc van Hal wrote: >> >>>On November 18, 2001 at 10:03:15, piet de hoop wrote: >>> >>>>The next position occured in Diep versus Rebel Century 4 round 6 of the Dutch >>>>Open Computer Chess 2001 >>>> >>>>[D] 8/6p1/3k3p/8/3K4/7P/8/8 b - - 0 43 >>>> >>>>The result was a draw, but for some computerprograms I tried, the evaluation was >>>>from -1 to -3.4 pawn. (Fritz 6.0, Crafty,Yace) >>>> >>>>I expect that programs equipped with the right pawntable base have no problem >>>>with this position, but i expected that programs without these tablebase while >>>>calculating over 30 plys or more should give the right evalution? >>>> >>>>Or do i miss something? >>>> >>>>Piet >>> >>>I think you mis something important here >>>Todays decent chessprograms use table bases so the chance that such a program >>>does not comeup with the right answer here is zero. >> >> >>Ok,but there are still decent programs that don't use the table bases. >> >>Anyway this is relatively an "easy ending", it is just counting. > >It may sound sarcastic, but I mean it serious. Could you give a specific rule >for such easy endings, where it is just counting? Perhaps i'm too optimistic, but when i should reached a position like this in my own games, i look for future positions which are deaddrawn. So for instance i don't play towards positions with only one pawn and two kings and possible wrong opposition. When i should have this position, the first thing i look for is how to fix the pawn at h3 with a black pawn on g5 and h4 and try to get the opposite king in a "zugzwang". And if this is possible i should get a easy win. If this isn't possible there normally isn't a clear win in these lines. In this position i cannot fix the pawn on h3 because the white king is on E4 and possible black pawns on h4 and g5 can be easily attacked. So if this plain already fails i quickly go back to the one pawn endings, and see that i älso cannot get a favourable "zugzwang" position. So it must be a draw. I cannot give an exact time i need for such evaluation, but i can do it in "tournament mode". I also think there are more ways to handle this ending. >>And when a program doesn't have the knowledge of this kind of ending, it just >>evaluates it as a positive ending way before in the search tree and it will >>exchange all the pieces just to go for it and discover in the end it is a draw! >>While in case it has knowledge about it, it won't exchange pieces and try a >>different route for a possible win. > >I think, many programs probe TBs rather deep in the search, so that such an >artivact will not happen too often. If it will happen, it might be an >indication, that the position was a draw already earlier. Anyway, also here I >would be interested if you could suggest a position, where you think, this may >happen. > >BTW. I gave this position to Yace without any TBs, and let it analyze for a >rather long time. The PVs are rather unreliable from the first occurence of an >attached "H" to the move, because of totally overloaded hashtables (I used 30 M >hash, K6-2 475):
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.