Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Hiarcs8 human-like qualities

Author: George Sobala

Date: 10:24:58 05/03/02

Go up one level in this thread


On May 03, 2002 at 10:21:24, Uri Blass wrote:

>
>Your reply suggest that you do not see his point.
>
>He claims that hiarcs was unable to find a positional plan in the nunn2 match
>that an average human has no problem to see.
>
>He did not claim that hiarcs was unable to find tactical ideas.
>
>
>I also do not see how +8 -3 against Fritz6 at 10 0 suggests that hiarcs is doing
>better at longer time control.
>
>This is a fantastic result in blitz time control.
>
>I guess that hiarcs was lucky because you need to be clearly better than Fritz7
>to expect 8-3 against Fritz6 and Hiarcs8 is not clearly better than Fritz7 based
>on other posts and it even lost 2-0 against yace at time control that is clearly
>slower than 10 0 and yace is clearly weaker than Fritz7.
>
>Uri

In that case I probably misunderstand the nature and purpose of the Nunn
positions. I thought they were a set of dynamic unclear positions with chances
for each side. I didn't realise there was a positional solution clear to an
average human player. Or do I still misunderstand?

As for the 10 0: I agree this is a fast time control. My earlier statement
related to bullet settings of 2 0 or 1 0 where Hiarcs8 suffers. A further
example of this is its performance on the Playchess server when run under two
different systems: on a virtual K6/2-450 it has an ELO around 2075, whilst on
the Duron/800 it is getting about 2450-2500. The Duron/800 calculates about 3.5x
as many nodes per second in the same position as the K6-2/450. The ELO
improvement of around 400 points seems excessive for a 3.5x speed difference.
Most of the games the engine has played on Playchess are at 5 2 or 10 2
settings.

George



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.