Will we see RYBKA 2.3.2 4CPU in cegt?

[ArmyBridge]

Hi, will CEGT or CCRL test RYBKA 2.3.2 in 4CPU ?? it would be very nice to see the level that this monster can reach in 4 CPU!! Regards

[Spock]

CCRL are currently testing it

It has to be said it has not got off to a strong start, but early results are notoriously unreliable anyway

[Werner]

Hi together,
thats fine if CCRL is testing Rybka with 4 CPUs.

I will test Rybka in CEGT with 4 CPUs when I have some opponents running on 4CPUs which can beat Rybka 2CPU

I am not interested in games where opponents can win 1 game out of 50.....

[Uri Blass]

Hi together,
thats fine if CCRL is testing Rybka with 4 CPUs.

I will test Rybka in CEGT with 4 CPUs when I have some opponents running on 4CPUs which can beat Rybka 2CPU

I am not interested in games where opponents can win 1 game out of 50.....

Maybe beating rybka 2 cpu is an harder task than beating rybka 4 cpu

I do not know if it is going to change but the rating of the 2 cpu is higher than the rating of the 4 cpu.

http://www.computerchess.org.uk/ccrl/40 ... t_all.html

Rybka 2.3.2 64-bit 2CPU 3124 +56 −53 76.3% −172.2 37.3% 118
70.1%
Rybka 2.3.2 64-bit 4CPU 3079 +57 −56 64.7% −89.7 48.9% 92



Uri

[Spock]

Yes, Werner's statement was a gross exaggeration

We have already tested Quad Rybka 2.2 4CPU, and there were plenty of opponents that were strong enough to play it - 7 opponents within 200 ELO and 10 within approx 250 ELO

The weekly update is running now and will be available in 3-4 hrs, with more Quad Rybka 2.3.2a games. At the moment the 2CPU rating is artificially high I think. This sort of thing can happen when there are less than a few hundred games. We have the same situation with Loop M1-T at the moment too. Maybe we'll see some movement in this update

[ArmyBridge]

Werner good point, I know.. Rybka 2CPU beat all of them on 4CPU and maybe still beat all of them on 6 or 8 CPU , but I like to see which level reach new Rybka 4CPU, I mean when Capa, Lasker or Kasparov dominated the chess world nobody wanted that they let play matches in their maximum force, so I Think that is justice that the fans of computer chess can see Rybka in his best, well is just my point of view I respect your point and your work that is very fine, Regards

[Spock]

On the website just now, you can see the 3 parings so far - and as you can see Quad Zanzibar came very close to Quad Rybka. And remember in WCCC, Zappa was a fairly close second behind Rybka. Remember these are 40/40 and may not be comparable to blitz results

Code: Select all

Rybka 2.3.2 64-bit 4CPU in CCRL 40/40, 2007-06-21
Ponder off, General books (up to 12 moves), 3-4-5 piece EGTB
Time control: Equivalent to 40 moves in 40 minutes on Athlon 64 X2 4600+ (2.4 GHz)
Tested by CCRL team, 2005-2007, http://computerchess.org.uk/ccrl/4040/

CCRL 40/40 main list:
based on 92 games: 37 wins, 10 losses and 45 draws
Score: 64.7%, Average opponent: −89.7, Draws: 48.9%


Pairwise results:
     Opponent                           ELO  Pure    Score                  LOS   Perf Pure
 - Zap!Chess Zanzibar 64-bit 4CPU       3046 3048  17.5-14.5  (+8 -5 =19)   86.0    -4
 - Hiarcs 11.1 4CPU                     2976 2984  21.0-9.0   (+15-3 =12)  100.0   +22
 - Loop M1-T 64-bit 4CPU                2942       21.0-9.0   (+14-2 =14)  100.0   -15

There will be more results in a few hours
All pairings will have 50 games when we have finished over the next couple of weeks

[Werner]

Yes, Werner's statement was a gross exaggeration

...I was waiting for this answer

[Spock]

Yes, Werner's statement was a gross exaggeration

...I was waiting for this answer

[Spock]

We now have:

Code: Select all

Rybka 2.3.2 64-bit 4CPU in CCRL 40/40, 2007-06-22
Ponder off, General books (up to 12 moves), 3-4-5 piece EGTB
Time control: Equivalent to 40 moves in 40 minutes on Athlon 64 X2 4600+ (2.4 GHz)
Tested by CCRL team, 2005-2007, http://computerchess.org.uk/ccrl/4040/

based on 138 games: 61 wins, 14 losses and 63 draws
Score: 67.0%, Average opponent: −106.4, Draws: 45.7%


Pairwise results:
     Opponent                           ELO  Pure    Score                  LOS   Perf Pure
 - Zap!Chess Zanzibar 64-bit 4CPU       3046 3048  17.5-14.5  (+8 -5 =19)   90.9    -6
 - Hiarcs 11.1 4CPU                     2976 2984  21.0-9.0   (+15-3 =12)  100.0   +20
 - Naum 2.1 64-bit 4CPU                 2955 2964  22.5-9.5   (+16-3 =13)  100.0    +2
 - Loop M1-T 64-bit 4CPU                2942       21.0-9.0   (+14-2 =14)  100.0   -17
 - Deep Fritz 10 4CPU                   2923 2942  10.5-3.5   (+8 -1 =5 )  100.0    +7

It is fair to say that we are disappointed so far - Quad Rybka 2.3.2a is not performing any better than Quad Rybka 2.2 it seems. But, it is foolhardy to draw any conclusions with so few games