CEGT - rating lists March 30th 2008

[Werner]

Hi all ,

our updated rating lists are now online and can be found under the attached links.
This week we all were waiting for Hiarcs 12. It came late - but not too late for our lists. See the first results here. The other new engines in our lists are Bright 0.3a and Garbochess 2.1.

40 / 120:
Our 40/120 Quad was not updated this week. Interim results as usual in our forum: http://husvankempen.de/nunn/phpBB2/viewforum.php?f=9
- with first Hiarcs 12 ShPV 4CPU games
- more ready games with Fruit 2.4 Beta A 4CPU
- and games with Zappa Mexico II, DS 11 and Glaurung 2.01 x64 all 4CPU

There is a close fight between Fruit and DS 11 now four the fourth position in the quad rating list with still a lot of games to come including matches for both against the new Hiarcs.

...and the result for the 3rd marathon match:

Code: Select all

CEGT Quad Extreme 40/400 repeated  2008 

1   Rybka 2.3.2a X64 4CPU  1½½½½½½1½½½½½1½½½½½½½½½1½½½½½101½½½½0½½½1½ 23.5/42 
2   Naum 3 X64 4CPU        0½½½½½½0½½½½½0½½½½½½½½½0½½½½½010½½½½1½½½0½ 18.5/42

A draw percentage of almost 80% but a better result than the Zappa versions against Rybka.

40 / 20:
This week we added more than 1900 games to our list. See more in our list "Games of the week". In total our 40/20 list is based now on 234.695 games!

New engines:
We started the matches with Hiarcs 12 2CPU with default settings as our time control isn´t one minute/move. We will later test Sharpen PV too.

Hiarcs 12 2CPU reached 2910 elos after 287 games. This is +38 elo above Hiarcs 11.1 2CPU - but we do not have enough games for a conclusion. We´ll see what happens with more games. The blitz games are much better!

Another good result we had with Bright 0.3a 2CPU: The engine has 35 elos more than version 2.0c after 544 games (2815)!

Garbochess 2.1 started with very good 2628 elos after 259 games. This is very close to Wildcat 7.

Updated engines:
We have updated results for some engines we tested first last week:

21 Fruit 2.4 Beta A x64 4CPU 2949 +29 -29 299 games (-21)
32 Fruit 2.4 Beta A x64 2CPU 2917 +19 -19 832 games (3)
132 SmarThink 1.10 Moscow 2759 +19 -19 865 games (+-0)
194 WildCat 8.0 2664 +20 -20 836 games (+2)
417 Rotor 0.2 2268 +39 -39 216 (now behind 0.1)

40 / 4:
Our blitz-list was updated too. We made over 3600 games and the list is based now on 291.877 games.

Some new interesting entries are:

18 Hiarcs 12 4CPU 2968 +26 -26 400 (here in front of Naum, Shredder and Fruit)
33 Hiarcs 12 2CPU 2940 +27 -27 390
55 Hiarcs 12 1CPU 2893 +52 -52 100 and

100 Bright 0.3a 2CPU 2821 +33 -33 300
141 Bright 0.3a 1CPU 2763 +28 -28 420

A big „Thank you“ to all testers as usual!

40/20: http://www.husvankempen.de/nunn/rating.htm
Blitz: http://www.husvankempen.de/nunn/blitz.htm
40/120: http://www.husvankempen.de/nunn/rating120.htm
Tester: http://www.husvankempen.de/nunn/testers/testers.htm
Games of the week: http://www.husvankempen.de/nunn/40_40%2 ... on/gow.JPG
Elo-comparison: http://www.husvankempen.de/nunn/Replay/ ... arison.htm

Werner
CEGT Team

[Yarget]

Hello Werner!

Thanks a lot for your great testwork. I will follow your Hiarcs 12 results with a great amount of interest and especially your blitztests. I wonder if there will be a small or big difference between your blitzresults and the one given at the official Hiarcs website:

http://www.hiarcs.com/hiarcs_games.htm

Best regards
Per

[Spock]

Hello Werner!
I wonder if there will be a small or big difference between your blitzresults and the one given at the official Hiarcs website:

http://www.hiarcs.com/hiarcs_games.htm

Best regards
Per

The conditions for that list are very favourable for Hiarcs. The choice of 32-bit not 64-bit means Rybka is compromised quite a lot just for starters.

[Yarget]

True Ray, the conditions aren't identical and therefore one should be careful to draw "big" conclusions. But even with these differences I am surprised to see the huge differences for these results:

Deep HIARCS 12 2 CPU v Rybka 2.3.2a UCI 2 CPU 51-49 (Hiarcs website)

Hiarcs 12 4 CPU - Rybka 2.3.2a x64 4 CPU 18-32 (CEGT tests)

I know..... Rybka 64 bit is stronger than the 32 bit version, the book effect and so on........ but still I didn't expect such a huge difference.

Nevermind, let's see when more testgames are done.

Best regards
Per

[Spock]

Well my 100 game FRC match was as follows (1CPU):

Hiarcs 12 vs Rybka 2.3.2 64-bit (36.0 - 64.0)

You will have seen yesterday the FRC update with 500 Hiarcs 12 games. In a few hours the final update with 1,000 games will be there. Most of the time the FRC results are pretty close to normal chess and CEGT's results so far seem to confirm that, although they need more games yet.

[Yarget]

You will have seen yesterday the FRC update with 500 Hiarcs 12 games.

Yes, I did see your results yesterday Ray. Your results so far seem to indicate that Hiarcs 12 is apx. 40 ELO points stronger than its predecessor. Not bad at all ofcourse but when I saw the results at the Hiarcs website (and these were the first results at all I saw with the new Hiarcs version) I expected more than 40 ELO-points.

Best regards
Per

[Spock]

Yes, I did see your results yesterday Ray. Your results so far seem to indicate that Hiarcs 12 is apx. 40 ELO points stronger than its predecessor. Not bad at all ofcourse...(snip)
Best regards
Per

Yes, that is very good. I'm not a programmer, but I think people underestimate how hugely difficult it is to get +50 ELO from the very strong engine that Hiarcs 11.1 already was, let alone any more than that. Yes we have seen some +80 and better recently (Naum, Shredder) but that is certainly the exception, not the rule, with top class engines. Mark continues to do an excellent job with Hiarcs. Of course different testng conditions may yleld different results - CEGT 40/120 will be the one to watch plus Sedat if he runs it

[Yarget]

To be more sure I just made some quick calculations. If my math is okay then the results at the Hiarcs website indicate that Hiarcs 12 is 72 ELO points stronger than its predecessor.

I hope that some of the ratinglists by CEGT and CCRL will confirm this improvement (that would ofcourse be great) but quite frankly I doubt it.......

Regards
Per

[Yarget]

Mark continues to do an excellent job with Hiarcs

I agree 100% with you Ray. Hiarcs remains one of the true topengines and I wish Mark good luck and hope that he will continues to improve Hiarcs. My comments so far shouldn't be considered as criticism, more a kind of disappointment that it appears (it is ofcourse still very early to make a final verdict) that Hiarcs like Naum and Shredder makes an +80 ELO improvement.

Regards
Per

[Yarget]

more a kind of disappointment that it appears (it is ofcourse still very early to make a final verdict) that Hiarcs like Naum and Shredder makes an +80 ELO improvement

Naturally I mean that Hiarcs won't achieve a +80 ELO improvement.