Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Spike-Rybka -From Expert To IM-

Author: Stuart Cracraft

Date: 09:38:06 12/28/05

Go up one level in this thread


On December 28, 2005 at 04:17:06, Rolf Tueschen wrote:

>On December 28, 2005 at 03:12:23, Vasik Rajlich wrote:
>
>>Interesting game. I didn't like .. h6 and .. g5 by Rybka in the opening (as I
>>suspect most human players won't), and there were a few moments in the endgame
>>where better resistance could have been put up - but these are relatively minor
>>points.
>>
>>The biggest issue was the incorrect sacrifice of the h6 pawn. The Rybka static
>>eval incorrectly gives the following position (and others of a similar type) as
>>roughly equal:
>>
>>[D] 8/1b2b1k1/pp1ppp2/nPq5/P1P1P3/2QRNPP1/8/1N2KB1r b - - 0 27
>>
>>I'll keep to myself my proposed solution. It's a tricky issue and won't be
>>handled until after the Rybka 1.2 release. If anybody wants to share here, I am
>>all ears :-)
>>
>>Vas
>
>
>I'm just an expert player and have plenty of weaknesses you as IM surely dont
>have. But I know how I would see the given position. And excuse me if I
>contradict an IM.
>
>1) Of course this position is good for Black. Period.
>
>2) Of course you cant exchange Queens in such a win/lose position! Period.
>
>3) Rybka is very good. Period.
>
>4) But the restrictions for chess in machine chess will always lead to such a
>nonsense, namely how Rybka destroyed a promissing position. The reason is simply
>the antique, historical, old, well known horizon effect. And there is no trick
>to lead a machine around this obstacle. Not by killer moves nor by nullmoves.
>Therefore - actually and for the next years - there is something wrong in your
>whole approach with the plan for 1.2.
>
>5) There is no trick to avoid the specifical chess Law in computerchess. The
>absolute Power of the Concrete in relation to the always very limited view of
>the machine due to horizon blindness. In the game against Spike every
>experienced player knows that he cant trade Queens in such an open attack
>position.
>
>6) Somewhere also YOU (as experienced and tricky IM) must hide the weakness to
>allow the telling brilliance of Rybka and that will be the key for your
>competitors, talking in engine-engine competition at least.

#4 of this is nonsense. The horizon effect has become less and less of
an issue since Berliner wrote about it as the march of technology has
advanced. I would think that Deep Thought and Deep Blue would have brought
that point to bear as singular extension was an algorithmic advance in
that area that was non-trivial, as implemented with the kinds of hardware
that helps it succeed.

Perhaps Bob can comment about his thoughts about horizon effect in modern
times.

Stuart




This page took 0.01 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.