Rybka 4.1

[NATIONAL12]

the latest testing on CEGT shows quite an increase for Rybka 4.1.



http://www.husvankempen.de/nunn/40_40%2 ... liste.html

[rbarreira]

Showing 9 elo right now. Is that a ~ 1 elo per month improvement rate or do I have the wrong release dates?

[mhalstern]

I think the elo gain is statistical randomness, as Vas stated that he did not change anything related to eval or search. This was a bug fix release, mostly related to stall bugs.

[LucenaTheLucid]

There 9 ELO improvement you are referring to is from the 6CPU version of Rybka 4 and the 4 CPU version of Rybka 4.1

[Dann Corbit]

Code: Select all

3 Deep Rybka 4.1 x64 4CPU 3253 22 22 612 71.5% 3094 41.0% 
6 Deep Rybka 4 x64 4CPU 3218 9 9 4170 73.9% 3037 36.5%    

Looks like 53-18=35 Elo to me.

[h1a8]

Why doesn't CCRL test Houdini with 4 cores? Why just test it with 1 core?

[M ANSARI]

Rybka 4.1 is only a bug fix with no ELO strength increase by way of evaluation improvement. Of course the stall bug was very serious and should improve ELO strength, especially in games where depth 13 or more are reached frequently.

[Jouni]

Let's see what happens in TCEC Elite match: Rybka 4 scored 16,5-23,6 = -60 ELO. Interesting match coming with super hardware!

Jouni

[jpqy]

Yes..i would like to see Rybka RC2 playing with his full strength against Houdini!

JP.

[Norm Pollock]

Can someone clear the confusion for me about "Rybka" and "Deep Rybka"? How are they the same or different?

Is "Deep Rybka 1CPU" the same as "Rybka"?