Reply
 
LinkBack Thread Tools Display Modes
  #1   Report Post  
Old October 9th 10, 09:10 AM posted to rec.games.chess.analysis,rec.games.chess.computer
external usenet poster
 
First recorded activity by ChessBanter: Jul 2010
Posts: 3
Default Rybka [5 sec] vs GetClub - Normal Level.

Sanny writes:

I want to conduct a match between Rybka [5 sec] vs Normal Level[4 min/
move].

....
Lets see how GetClub plays against the strong Engines.


Lets see how GetClub plays against very weak engines first. Here is a
game I played between my Novag Constellation 3.6mhz (see
http://www.ismenio.com/chess_novag_c...lation_36.html) and getclub
beginner level:

[Event "?"]
[Site "?"]
[Date "2010.10.09"]
[Round "-"]
[White "Novag Constellation 3.6mhz, 40 moves in 5 minutes"]
[Black "GetClub beginner level (average 30s per move), 2.6GHz Core2"]
[Result "1-0"]

1. e4 e5 2. Nf3 Nc6 3. Nc3 Nf6 4. Bb5 Bb4 5. O-O O-O 6. a3 Bxc3 7. dxc3 d6
8. Bxc6 bxc6 9. Bg5 Bg4 10. Qd3 Bxf3 11. Qxf3 Qb8 12. Bxf6 Qxb2 13. Qg4 g6
14. Qg5 Qxc3 15. Qh6 {Black resigns}
1-0

Normally a single game doesn't mean that much, but in this case I think
it shows more than enough, because of the following:

* The getclub computer is at least 10000 times faster than the Novag
computer.

* The getclub computer has 250000 times more memory than the Novag
computer.

* The getclub program used on average 4 times more thinking time than
the Novag program.

* A program playing 11... Qb8 and 12... Qxb2 has serious problems.

* At moves 13 and 14, getclub didn't see the forced mate even though it
showed a search depth of more than 20.

Until getclub gets a working search function that can find basic
tactical combinations, there is no point trying to measure its strength
by playing games versus strong chess engines.

A very modest requirement on a working search function is that it should
score at least 200 of 300 on the WAC test suite, at 5s thinking time on
a standard PC. I think the last time this was tested, getclub only
scored about 25% or so.

--
Peter Osterlund -
http://web.comhem.se/petero2home
  #2   Report Post  
Old October 9th 10, 06:16 PM posted to rec.games.chess.computer
external usenet poster
 
First recorded activity by ChessBanter: Jan 2006
Posts: 470
Default Rybka [5 sec] vs GetClub - Normal Level.

On Oct 9, 1:10*am, Peter Osterlund wrote:

A very modest requirement on a working search function is that it should
score at least 200 of 300 on the WAC test suite, at 5s thinking time on
a standard PC. I think the last time this was tested, getclub only
scored about 25% or so.


That would be a VERY modest requirement. The very first version of my
engine that could accept positions (meaning you could run test suites
on it) and was rated around 1650 got 244 out of 300 at 5s per position
-- and that was after I had been working on it for only about four
months.

jm
  #3   Report Post  
Old October 10th 10, 03:13 PM posted to rec.games.chess.computer
external usenet poster
 
First recorded activity by ChessBanter: Jul 2006
Posts: 1,015
Default Rybka [5 sec] vs GetClub - Normal Level.

On 09/10/2010 09:10, Peter Osterlund wrote:
writes:

I want to conduct a match between Rybka [5 sec] vs Normal Level[4 min/
move].

...
Lets see how GetClub plays against the strong Engines.


Lets see how GetClub plays against very weak engines first. Here is a
game I played between my Novag Constellation 3.6mhz (see
http://www.ismenio.com/chess_novag_c...lation_36.html) and getclub
beginner level:

[Event "?"]
[Site "?"]
[Date "2010.10.09"]
[Round "-"]
[White "Novag Constellation 3.6mhz, 40 moves in 5 minutes"]
[Black "GetClub beginner level (average 30s per move), 2.6GHz Core2"]
[Result "1-0"]

1. e4 e5 2. Nf3 Nc6 3. Nc3 Nf6 4. Bb5 Bb4 5. O-O O-O 6. a3 Bxc3 7. dxc3 d6
8. Bxc6 bxc6 9. Bg5 Bg4 10. Qd3 Bxf3 11. Qxf3 Qb8 12. Bxf6 Qxb2 13. Qg4 g6
14. Qg5 Qxc3 15. Qh6 {Black resigns}
1-0

Normally a single game doesn't mean that much, but in this case I think
it shows more than enough, because of the following:

* The getclub computer is at least 10000 times faster than the Novag
computer.

* The getclub computer has 250000 times more memory than the Novag
computer.

* The getclub program used on average 4 times more thinking time than
the Novag program.

* A program playing 11... Qb8 and 12... Qxb2 has serious problems.


Erm. Yes. Even the basic evaluation function should get the right move
and avoid moves that get the kings castled position ripped apart instantly.

11. Qb8 ranks 15/22

But 12. ... Qxb2 is the best yet! At 22/22 no other move is worse.
It sets up a forced mate that should be well inside most engines search
horizon. Certainly the Novag had no trouble spotting it!

12. .. gxf6 is so obvious that it is incredible that any engine could
play anything else. Basic material eval gives the right answer at ply 1
and nothing else should be able to supplant it!!!

* At moves 13 and 14, getclub didn't see the forced mate even though it
showed a search depth of more than 20.

Until getclub gets a working search function that can find basic
tactical combinations, there is no point trying to measure its strength
by playing games versus strong chess engines.

A very modest requirement on a working search function is that it should
score at least 200 of 300 on the WAC test suite, at 5s thinking time on
a standard PC. I think the last time this was tested, getclub only
scored about 25% or so.

By standard PC do you mean one running at 4.7MHz?

The ancient VAX engine I am resurrecting from old Pascal sourcecode is
now at a stage where on a 3GHz box it will score about 200 on WAC at an
average of 5s/move. And it scores 188/300 @ 0.3s/move.

Its performance in terms of improved search depth with increasing time
does not scale anything like as well as modern engines that use much
more advanced aggresive pruning techniques.

Regards,
Martin Brown
Reply
Thread Tools
Display Modes

Posting Rules

Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
GetClub Easy Level vs Rybka Sanny rec.games.chess.politics (Chess Politics) 7 May 27th 08 01:03 PM
GetClub Easy Level vs Rybka Sanny rec.games.chess.misc (Chess General) 10 May 27th 08 01:03 PM


All times are GMT +1. The time now is 10:23 AM.

Powered by vBulletin® Copyright ©2000 - 2019, Jelsoft Enterprises Ltd.
Copyright 2004-2019 ChessBanter.
The comments are property of their posters.
 

About Us

"It's about Chess"

 

Copyright © 2017