Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: KasparovChess.com: A few questions for ChessBase

Author: Robert Hyatt

Date: 15:03:56 09/07/01

Go up one level in this thread


On September 07, 2001 at 16:27:40, Mike S. wrote:

>On September 07, 2001 at 14:36:08, Robert Hyatt wrote:
>
>>This is easier than you think.  Remember how the fritz programs were killing
>>chess tiger on ICC?  playing 1. h4 which would eliminate the normal opening
>>book preparation.  Nemeth has shown that this can be reproduced with other
>>similar "short opening line" ideas. (...)
>
>I don't know those 1.h4 games, but I don't think that Kramnik can apply
>sacrificial, tricky opening ideas. This sounds way to risky to me. I assume, an
>8 CPU machine does not necessarily play the same moves like a dual just given
>more time. He can't risk that.

The risk is minimal.  If the program doesn't understand the trojan horse
attack, it will fall for it regardless of how fast the hardware is.  If it
doesn't understand how to play against 1. h4, then faster hardware isn't
going to help.  And if Kramnik takes a 1ghz single-cpu machine, and tries
various moves on it giving it up to 8x the normal search time, he can pretty
well predict how it is going to play, and set up some serious "sucker-play"
positions.



>
>Such games can be reproduced similar, but I don't think it's possible at the 1st
>try always. Probably there's a number of failing attempts, until a final version
>of such a game crystallizes. Typically, such games are a tightrope walk for some
>moves, with the program missing the best defenses, aren't they?


Depends.  Many are just lost.  IE the trojan horse attack is a trivial win
for a human once the program 'bites'.





>
>But a match game is only one try each...
>
>Regards,
>M.Scheidl



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.