Author: Kurt Utzinger
Date: 09:40:45 08/05/03
Go up one level in this thread
On August 05, 2003 at 10:01:52, zeidoun wrote:
>Yes Kurt, the book choice has an important issue in the games result. In my
>little experience, i've already noticed this and i've submitted many questions
>and also many responses in this forum. But the fact is that i'm not Kurt
>Utzinger nor Harald Faber, I'm just Zied Haddad from the chesslovers.
>As i mailed to u i've developed new settings for shredder 7.04 that fit not only
>the blitz time controls but also rapid time controls like 25mn games. This
>superb setting is shredder 7.04_Attack1.2, u will find it in my homepage at:
>http://www.chesslovers.tk/ u have then to check the english page (the other
>pages are for the moment no longer up-dated). You enter then into the tournament
>page u will find the details there. I haven't yet up-dated my page to show you
>how incredible is the strength of this new setting for shredder in 25mn games
>but also in blitz games.
>It won clearly against all the top 10 engines in individual matches (making for
>example 70% of the points against Fritz 8, where shredder 7.04 with his
>beginning settings wins only with a games up). But i've tried an overall
>simultaneous match with the top 25 commercial rated programs (exept the king),
>and i'm surprised with a poor peformance against some oldies programs (shredder
>is rated normally more than 150 points of ELO), this poor performance is due to
>the choice of the opening.
>In my test, i used 5 modified positions from the Nunn1 TestSet. The position
>from which all the programs begin to play rises after 5 or 9 moves. All these
>positions are with equal chances for both parts, none of these positions are
>played with an anti-computer style (for example we choose to only develop our
>paws and our pieces in the third line). My purpose was to ovoid the influence of
>book opening. And all the positions where played with black and white to show
>the real strength of the engines.
>I previously posted a question dealing with the learning option that's delivered
>with shredder or hiarcs, but none has responded to me, i think that's because
>none has thought about this point of view. Wait for the next upate of my site to
>check the games and the rating of this new shredder setting. I think that this
>prooves that not only chessmaster but also shredder could be improved, and i
>think that this is a good point for these programs, because their conceptors
>could give us in the next future a more powerful product. <it not the case of
>fritz that reached out his adult age where the development of newer algorithms
>will be more and more difficult. I also think that the next development of fritz
>could be may be the search for new settings.
Hi Zeidoun
I have followed your tests with Shredder 7 and 7.04 published on your
homepage but never made an attempt to try the Shredder 7.04 Attack1.2
settings. The main reason for this is that I never test old vs new
version of the same engine as I believe [this is just my perhaps wrong
opinion] this does not prove much [if anything at all]. And furthermore
- as I do not like blitz games - I did not want to make own tests with
other settings of Shredder 7.04 that are not based on games at longer
time controls [40'/40 and higher]. The lowest test level I use to get a
very first and fast impression is rapid chess 25m+10s. However, I do not
think that your testing is worthless. And why should it not be possible
to improve the very strong Shredder 7.04 with some amended settings? I
have however nothing heard so far abount stronger settings for this engine
at longer time controls and it would therefore be of great interest if
Shredder 7.04_Attack 1.2 could then confirm the good results too. By the
way, there are some other things that are still not [enough] tested:
influence of hash table size, book type, book learning, position learning,
use of EGTB and so on. And this is the reason why you don't get many [if
any at all] answers to such questions. It is still my dream that chess
computer testers all over the world would join to discuss and find a way
of systematically testing chess programs.
Kind regards
Kurt
http://www.utzingerkurt.com
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.