Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Rybka 1.0 Beta 32-bit Test (40'/40) After 500 games (Match vs Loop L

Author: George Tsavdaris

Date: 13:28:58 12/19/05

Go up one level in this thread


On December 19, 2005 at 15:06:27, Kurt Utzinger wrote:

>In our 5moves 40'/40 test with matches of 50 games each,
>Rybka 1.0 Beta 32-bit has so far beaten all other engines.
>It was therefore no surprise that Rybka also won vs Loop
>List 600 by 35,0-15,0 (70 %).

 Many thanks of course for all your testing! It's nice i think to tell you that
from time to time, although i mean it every time.........
 I've observed many of your games and as i've already noticed, Rybka seem to
know what it is doing at every move. Its moves seem to have a deep purpose that
the opposition doesn't understand.....That's why they lose.......Of course the
way it does this(understabding position better than others) is not perfect
otherwise it would be invincible!

>And here we should like to
>give a warning: in the first 10 games Loop List got a score
>of 55 % (5,5/10) but in the end this score dropped to 30 %
>which means that in the following 40 games Loop List only
>made 9,5/40 (27,5 %) ... just in case for people who still
>thinks that 10 games are enough -:)

A tricky moment:

And here i would like to give a warning: in the first 50 games Loop List got a
score of 30 % (15,0/50.0) but in the end after 200 games this score increased to
48 % which means that in the following 150 games Loop List made 81/150 (54 %)
... just in case for people who still thinks that 50 games are enough -:)

:-):-)

10 games are enough for people interested into a specific error margin.
50 games are enough for people interested into another smaller error margin.
200 games are enough for people interested into a more strict error margin.




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.