Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: breaking blocked positions

Author: Dan Honeycutt

Date: 18:38:41 07/09/04

Go up one level in this thread


On July 09, 2004 at 19:43:01, Uri Blass wrote:

>On July 09, 2004 at 17:12:00, Dan Honeycutt wrote:
>
>>On July 09, 2004 at 13:24:16, Duncan Roberts wrote:
>>
>>>sometimes humans to get a draw against computers, block the position
>>>leaving the computer wandering about aimlessly doing nothing.
>>>
>>>
>>>
>>>why cannot the program tell the computer if nothing is happenning (ie 25 moves
>>>have gone (1)without any captures (2) the score is virtually the same
>>>(3)none of the computers pieces have gone over into the last 3 ranks and vice
>>>versa)  then the computer should send over either a bishop, queen or rook into
>>>the last 3 ranks if this can be done without losing more than a 1/3 pawn.
>>>
>>>
>>>This should break open the position.
>>>
>>>
>>>
>>>Duncan
>>
>>
>>Should break it open but might well cost you the game.  A problem with the
>>alpha-beta search is that you only know the score of one move - the best one. So
>>you really don't know if your liberating move is only going to set you back 1/3
>>pawn, it could be a total disaster.
>>
>>Dan H.
>
>You do not need a score for other moves and you only need evaluation that tells
>the computer that 50 plies without capture together with some other conditions
>is bad for itself.
>
>The question if it is a good idea is another question and I do not think that
>it is productive against computers and the interesting game for me is comp-comp.
>
>Uri
>
>Uri


If a computer is happy with it's score it has no incentive to make progress in
the game until something happens - it finds something better, the 50 move or 3
rep rule approaches, the opponent forces the issue, etc.  I took the the
original poster's question to be "when you see a lack of progress, why not prod
it along by playing a liberating (albeit slightly inferior) move".  My reply was
just that that can be a very risky proposition.

Dan H.



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.