Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: rybka pawn pushes. can new versions of engines handle it?

Author: Robert Hyatt

Date: 15:51:57 02/25/06

Go up one level in this thread


On February 25, 2006 at 11:01:13, Uri Blass wrote:

>On February 25, 2006 at 04:28:52, Joseph Ciarrochi wrote:
>
>>I just watched yet another game involving the famed rybka pawn push....Glauring
>>did not seem to recognize the danger of the passed pawn till it got to the 6th
>>row, wheras rybka seemed to recognize it immiediately..
>>
>>I have two questions regarding this advantage rybka seems to have over all
>>others
>>
>>1) Once you see this error in chess engines (not recognizing the danger of a
>>passed pawn early enough), how easy is it to program your engine to correct the
>>weakness. Is this rybka's major "trick" and once engines adjust to it, will
>>rybka be pulled back to equal with the other engines.
>
>You first need to define the weakness.
>If you do not know to define what rybka does in the evaluation in english
>language you have no chance to program it.
>
>>
>>2) this leads me to a more general question. are chess engines getting better
>>against humans, or are they only getting better against other engines. This
>>rybka pawn push......would super GM's fall for it. Did rybka merely expose a
>>latent weakness that exists only in other engines but not in humans (e.g.,
>>humans are usually pretty good at seeing long term pawn push plans)
>
>I do not agree that humans are usually pretty good at seeing long term push
>plans.
>Maybe super GM are good at it but humans generally are not good at it and I
>guess that I could be a better chess player if I had the knowledge of rybka
>about it.
>
>Uri

Humans can do this far faster and easier than computers do.  I don't have much
trouble looking at a passed pawn, at the squares in front of it, and then
focusing solely on the question "can I advance this pawn, can I break every
blockade my opponent sets, can I control every square along the way to
promotion, etc...???"

It isn't always easy, but then I am only looking at trying to see if I can
advance the pawn or if it is going to get permanently stuck somewhere down the
file and be worthless for the effort expended to get it that far...

It just isn't so easy to program that and not have it kill the speed of
evaluation for all the positions where that is not so important...




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.