Author: pavel
Date: 18:31:42 04/29/05
Go up one level in this thread
On April 29, 2005 at 01:52:45, Kurt Utzinger wrote:
> Hi Pavel
> I can't no longer remember about the purpose of
> your testing these two engines. And what are the
> playings conditions. Many thanks for posting
> these details again.
> Regards
> Kurt
Hi Kurt,
Sorry for late reply. The pupose of the testing was two fold.
1) To see if a large amount of games are played between two engines, can the
weaker engine match the strength eventually of the stronger opponent, with the
help of positional learning. The reason Yace was chosen (the weaker engine),
because it has oe of the best positional learning among the free engines. The
reason Aristarch was choosen, other than that I have not done much testing with
it myself, it is stronger than YACE but not by a very wide margin.
2) To see how many games are needed to play to reach "0" error margin in "rating
differance". In this case Elostat was used to calculate rating and error margin.
There might be a mathematical approach in calculating the number of games needed
to reach "0" error margin. But I opted to go for the 'hard way'.
Besides both purpose of the test compliments each other, one can't be reached
without the other, (maybe partially).
However, I am not too excited to run 100,000s games. So I will see how far I can
go. Currently about 6000+ games has been played. Target at the moment is 10,000
games.
The state of the playing conditions are really sorry looking, but I had no other
options.
Chessbase GUI
Own Book
Learn on (Yace Book and Positional Learn, Aristarch only supports book learn)
TB ON
Time Control: 2min/game
Ponder off
Hash: 50+
Hardware: PIV 2.66/760mb Ram
----------------------------------------
Cheers,
pavs
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.