Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Analysing while retracting moves

Author: Robert Hyatt

Date: 09:26:41 11/27/01

Go up one level in this thread


On November 27, 2001 at 01:11:52, Uri Blass wrote:

>On November 26, 2001 at 22:58:13, Robert Hyatt wrote:
>
>>On November 26, 2001 at 19:34:48, Tom Kerrigan wrote:
>>
>>>On November 26, 2001 at 15:12:59, Robert Hyatt wrote:
>>>
>>>>If it were one move, you would be right.  But it typically isn't.  It is
>>>>_many_ moves.  Say 10 moves or 20 plies.  And the score doesn't get backed up
>>>>the the _real_ problem move at all...
>>>
>>>Ah, here's where we fundamentally disagree. Do you have any evidence that
>>>"typically" mistakes can only be found with 20+ plies of search? 20 sounds to me
>>>like a number that was pulled from somewhere unpleasant.
>>>
>>>-Tom
>>
>>
>>Try some games.  Like the famous Shirov sacrifice, and see where the program
>>thinks white goes wrong.  Those are the kinds of positions that are important
>>to the discussion.  If the problem is only detected one move late, that can
>>be fixed by 10x more search time.  It is the ones that are much deeper that
>>cause the problems...
>
>It may be interesting to give yace some hours per move and let it analyze when
>it goes backward.
>
>I believe that it has a chance to find the real mistake in these conditions even
>in that game.
>
>There are also cases when there is a long forced line also from computer point
>of view[it is the case in part of the nolot positions(for example nolot number
>2) and in part of the studies]
>
>Uri


Have it analyze _any_ game where there is a long-term kingside attack.  It
will continually suggest different moves at various points in the game,
yet those moves won't change a thing.  Information passed backward up a
tree is a tiny percentage of the information needed to _really_ understand
what is happening.

If you like answers with a lot of serendipity, you got it.  :)



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.