Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka 3 CEGT results
1 2 Previous Next  
Parent - By Venator (Silver) Date 2008-08-09 13:09
Once again thanks, Heinz. These are incredible results, even better than I had myself.....
Parent - By ThatsIt (*) Date 2008-08-11 14:59
Parent - - By Patricio (***) Date 2008-08-09 15:05
Heinz,

First of all, thank you for your excellent Job.

I'd like to know what the time controls of this games are.

Best regards,

                             Patricio.
Parent - - By Heinz van Kempen (***) Date 2008-08-09 15:10 Edited 2008-08-09 15:23
Hola Patricio :-),

40/4 repeated. Fortunately Rybka is winning a lot of short games, so we proceed quickly.

Seems that I posted general conditions only in CEGT Forum and not here. So here they are:

Conditions:
CEGT Blitz 40/4 repeated
128 MB hash each engine
5 men EGTB
test suite with one hundred games from Super GM tournament practice 2007/2008 collected by me, including the most popular openings currently, restricted to 8 moves, each position was tested by me beforehand in order to give equal chances to both sides, same opening is played with colors reversed for more precision
used hardware is 4 x Core 2 Quad E6600 (5 men) - 4 GB RAM

I´m using five machines currently 24/7. Four quads and a dual core machine.

No crashes.

Best Regards
Heinz
Parent - - By Felix Kling (Gold) Date 2008-08-09 15:16
". Fortunately Rybka is winning a lot of short games, so we proceed quickly"

hehe :)
Parent - - By Heinz van Kempen (***) Date 2008-08-10 10:02
Hi all  :) ,

here a last overview from my side. I have started 40/120 just now.

For all Rybka versions the same set was used with colors reversed additionally, that means all played there 200 games matches with exactly the same opening lines.

The Blitz rating list by Gerhard later today will surely look differently from this results (some more from others are posted in the CEGT Forum), because Gerhard has a lot of other games to include for Rybka 3 and others run by other testers.

[code]Individual statistics:

1 Rybka 3 x64 4CPU          : 3263  1000 (+706,=235,- 59), 82.3 %

Deep Shredder 11 x64 4CPU     : 200 (+149,= 45,-  6), 85.8 %
Zappa Mexico II x64 4CPU      : 200 (+130,= 60,- 10), 80.0 %
Deep Fritz 10.1 4CPU          : 200 (+143,= 41,- 16), 81.8 %
Hiarcs 12 4CPU                : 200 (+162,= 29,-  9), 88.2 %
Naum 3.1 x64 4CPU             : 200 (+122,= 60,- 18), 76.0 %

2 Rybka 3 Dynamic x64 4CPU  : 3240  1000 (+680,=247,- 73), 80.3 %

Deep Shredder 11 x64 4CPU     : 200 (+141,= 50,-  9), 83.0 %
Zappa Mexico II x64 4CPU      : 200 (+113,= 67,- 20), 73.2 %
Deep Fritz 10.1 4CPU          : 200 (+134,= 52,- 14), 80.0 %
Hiarcs 12 4CPU                : 200 (+165,= 25,- 10), 88.8 %
Naum 3.1 x64 4CPU             : 200 (+127,= 53,- 20), 76.8 %

3 Rybka 3 Human x64 4CPU    : 3224  1000 (+654,=268,- 78), 78.8 %

Deep Shredder 11 x64 4CPU     : 200 (+137,= 48,- 15), 80.5 %
Zappa Mexico II x64 4CPU      : 200 (+123,= 58,- 19), 76.0 %
Deep Fritz 10.1 4CPU          : 200 (+143,= 38,- 19), 81.0 %
Hiarcs 12 4CPU                : 200 (+129,= 53,- 18), 77.8 %
Naum 3.1 x64 4CPU             : 200 (+122,= 71,-  7), 78.8 %

4 Rybka 3 x64 2CPU          : 3208  1049 (+734,=238,- 77), 81.3 %

Deep Shredder 11 x64 2CPU     : 200 (+146,= 40,- 14), 83.0 %
Zappa Mexico II x64 2CPU      : 200 (+128,= 56,- 16), 78.0 %
Fruit 2.4 Beta A x64 2CPU     :  62 (+ 54,=  7,-  1), 92.7 %
Hiarcs 12 2CPU                : 187 (+123,= 50,- 14), 79.1 %
Naum 3.1 x64 2CPU             : 200 (+152,= 39,-  9), 85.8 %
Deep Fritz 10.1 2CPU          : 200 (+131,= 46,- 23), 77.0 %

5 Rybka 2.3.2a x64 4CPU     : 3090  1950 (+1043,=737,-170), 72.4 %

6 Rybka 2.3.2a x64 2CPU     : 3056  2694 (+1426,=1000,-268), 71.5 %[/code]

Best Regards
Heinz
Parent - - By Venator (Silver) Date 2008-08-10 10:19
So far the Quad version of Rybka 3 has +173 elo compared to Quad Rybka 2.3.2a. That is even better than Larry's optimistic guess!
Parent - By Felix Kling (Gold) Date 2008-08-10 10:27
Seeing those numbers makes me smile :)
Parent - - By buffos (Silver) Date 2008-08-10 10:27
and the main "problem" is how can Rybka 4 even achieve even half of that progress !
Parent - By Felix Kling (Gold) Date 2008-08-10 10:30
well, that's no problem. Look at the Rybka games in mainz for example. It was far away from perfect play- quite a lot room for improvement. Some things like getting space and pawns that help rooks to enter the 7th rank and so on could be improved quite a bit.
Parent - - By Heinz van Kempen (***) Date 2008-08-10 10:29 Edited 2008-08-10 10:33
Hi Jeroen :-),

yes, much more than could be realistically expected.

The used opening set I configured from CBM and TWIC games from current Super GM practice is availabe for download here.

http://www.husvankempen.de/nunn/Replay/download40120.htm

Quad games for CEGT 40/120 repeated (real time control, not adapted to weaker hardware like CEGT and CCRL usually do) will use the same set.
Tests with own books and this time control will be given as soon as your book will become available.

As there is currently a lot of discussion how Rybka 3 would perform with long time controls I will give daily an update (at least over the next week) with this games here in the Forum, so that you can replay them. Will also upload the pgn with comments for this games regularly to our server. This will start tonight with first 8 to 12 games. We will need many weeks for an estimation, but at least can study the games, errors, highlights, etc. before we are going from chess to numbers (stats).

Best Regards
Heinz
Parent - - By Venator (Silver) Date 2008-08-10 10:38
Hi Heinz,

Thanks for all the updates and the download! When will the official CEGT update be published, maybe this evening? I need it for my article in the Dutch magazine 'Computerschaak' :-).

Yes, it is interesting to see if the same results can be achieved at slower time controls. I have played a few matches game in 30, game in 60 and game in 90 and they suggest that scores do not differ too much. But game samples were rather small, so I am looking forward to the CEGT 40/20 and 40/120 results!

Kind regards, Jeroen

PS  Of course I am very much interested in the results with my book as well!
Parent - By Heinz van Kempen (***) Date 2008-08-10 10:41
Hi Jeroen :-),

I guess this evening. I do not know if all testers sent the games to Gerhard already or if he prefers to watch the Olympics before starting to calculate :-).

Best Wishes
Heinz
Parent - - By Heinz van Kempen (***) Date 2008-08-10 11:16
Hi Jeroen :-),

you can report in 'Computerschaak' which I read in the 90´s, together with "Modul", "CSS" and others.

http://www.husvankempen.de/nunn/40_4_Ratinglist/40_4_AllVersion/rangliste.html

Rybkaish (Rybkaesk) Regards :-)
Heinz
Parent - By Venator (Silver) Date 2008-08-10 12:34
Superb! Great job by the CEGT, as ever! Thanks :-) :-)
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka 3 CEGT results
1 2 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill