Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A couple positions from today's ICC tourney games - corrected

Author: Howard Exner

Date: 03:34:22 02/02/00

Go up one level in this thread


On February 02, 2000 at 06:29:43, Howard Exner wrote:

>On February 01, 2000 at 17:26:40, Jeremiah Penery wrote:
>
>>On February 01, 2000 at 05:55:07, Howard Exner wrote:
>>
>>>On February 01, 2000 at 02:39:46, Jeremiah Penery wrote:
>>>
>>>>Thanks for all your analysis!  I think I see now that white was lost anyway,
>>>>though I'm not 100% convinced yet (only 99% :).  In any case, it would be really
>>>>difficult for either side (white to hold the draw, if it existed, and for black
>>>>to find the win, if white did optimal play).
>>>
>>>These positions are always lots of fun to hash out with other chess players.
>>
>>Yes!  I agree. :)
>>
>>>Eventually we seem to get to the truth of a position with the help of our
>>>computer programs ( at least in my case).
>>
>>I couldn't do it without the computer, for sure.  I make simple tactical errors,
>>that the computer can correct for me, while I can try to decide on the strategy.
>>
>>>We sort of go into correspondence
>>>analysis mode (using a computer to boot) while the chess program is bound to the
>>>limits of the time control it played under. So I agree that this and other
>>>positions are difficult for machines when faced with optimal play. Would they
>>>play h4 here and follow up correctly? Would they play the Rg4+ move and
>>>the Ke5 response!(that you found) from that other interesting thread?
>>
>>Before very recently, I never truly appreciated the difficulty in such analysis.
>> I had done some much simpler analyses, but I never had anyone to help me find
>>the truth of a position.  Of course, I always made mistakes.  Before, when I saw
>>that, say, a GM's analysis had an error, I was really amazed that they could
>>miss such a thing.  Now, I realize just how difficult it is to make sure all the
>>moves in the analysis is optimal, and that alternate lines are provided.  In
>>some positions, it is really difficult to do.
>>
>>Thanks for helping me find the truth in this position. :)
>>
>>>Mind you many of todays programs are playing very nice endgame moves when
>>>compared to say 3-5 years ago.
>>
>>True, but it still seems to be one of the major weak points of computers.  There
>>are still too many pieces for them to be able to use TBs, and they just have no
>>plan on what they need to do.  In the middlegame this seems to work Ok, but in
>>the endgame they need a plan, or else they too often simply shuffle pieces.
>
>This idea of needing a plan is indeed a stumbling block in endgames. It's
>tied in with pattern recognition. Here's an example:
>
>This is a sequence of three positions from the same game
>starting with the final one first and working backwards.
>
>1. Humans and computers know this is an easy win for black.
>Computers by calculating, humans at a glance through pattern recognition.
>
>[D]8/5p2/4p1p1/4P2p/5P1P/kP4P1/8/1K6 b - - bm Kxb3;
>
>2.In this position humans can also see quickly that
>the queen side pawns can be exchanged leading to the
>first position.
>
>[D]8/5p2/p3p1p1/1p2P2p/5P1P/kP4P1/P7/1K6 b - - bm a6a5;
>
>3. This initial position will lead to the other two losing positions
>if white plays Bxc5? Humans, knowing the other two patterns
>will avoid the exchange and instead consider moves like Be3
>or Be1 but computers will have to calculate this one
>to great lengths to avoid losing. How many programs will avoid
>Bxc5?
>
>[D]8/5p2/p3p1p1/1pbkP2p/5P1P/1P4P1/P3KB2/8 w - - am Bxc5

Sorry, the diagrams are corrected with black to move for the first
two positions



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.