Nirvanachess 2.2 Release

[cetormenter]

Hello all,

Nirvanchess 2.2 has been released.

More details can be found at http://nirvanachess.blogspot.com/2015/1 ... lease.html

The biggest changes in this version are bug fixes to both single and smp searches. Smp should also scale well beyond 4 cores.

Overall strength increase is somewhere in the range of 50 elo.

[TShackel]

Thanks Thomas, and congrats on the progress. Any chance of getting multivariation mode into a future release?

Thanks.

Tim.

[Graham Banks]

Thanks Thomas. Great engine!

[cetormenter]

Perhaps. There are already quite a few engines that are stronger than Nirvana that have this feature though. I am not aware that too many people use Nirvana for analysis.

[Dann Corbit]

Perhaps. There are already quite a few engines that are stronger than Nirvana that have this feature though. I am not aware that too many people use Nirvana for analysis.

Considering how strong Nirvana is becoming, I suspect that more and more people will be interested in this feature.

[IWB]

Hello Thomas,

... Overall strength increase is somewhere in the range of 50 elo.

On your Blog you wrote "Overall strength gain ~50 elo after 5000 games against 2.0a at 60"+0.05". Is there any strength increase expected to Version 2.1c?

Thx
Ingo

[Sylwy]

On your Blog you wrote "Overall strength gain ~50 elo after 5000 games against 2.0a at 60"+0.05". Is there any strength increase expected to Version 2.1c?

2.1c......................3032 Elo CCRL 4/40
2.0a......................2973 Elo CCRL 4/40
---------------------------------
+59 Elo gain

2.2 face to 2.0a..........+50 Elo gain

=> 9 Elo points downgrade !

?????

SilvianR

NR :

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

[cetormenter]

Hello Thomas,

... Overall strength increase is somewhere in the range of 50 elo.

On your Blog you wrote "Overall strength gain ~50 elo after 5000 games against 2.0a at 60"+0.05". Is there any strength increase expected to Version 2.1c?

Thx
Ingo

I don't know what you are talking about .

I simply forgot to update that part from my copy and paste from the previous post. The ~50 elo gain is compared to 2.1c. Obviously this is in self play at quick time controls so I expect the gain on ratings list to be smaller. Probably closer to 20 elo or so.

Here is a short run from a gauntlet I am currently running:

Code: Select all

Rank Name         Elo    +    - games score oppo. draws 
   1 Stockfish    288   35   35   490   89%   -42   18% 
   2 Gull         268   34   34   494   87%   -42   17% 
   3 Senpai        18   31   31   362   61%   -57   29% 
   4 NirvanaDEV     0   22   22   781   43%    63   25% 
   5 Texel        -11   27   27   490   54%   -42   22% 
   6 Protector    -40   26   26   492   50%   -42   34% 
   7 Nirvana2.1   -57   12   12  2534   36%    57   26% 
   8 Naum         -59   26   26   493   47%   -42   30% 
   9 Hannibal     -64   26   26   494   47%   -42   31% 

[IWB]

I simply forgot to update that part from my copy and paste from the previous post. The ~50 elo gain is compared to 2.1c.

PERFECT. Nirvanachess will be my next thing to test.

Thx
Ingo

[mar]

Obviously this is in self play at quick time controls so I expect the gain on ratings list to be smaller. Probably closer to 20 elo or so.

Congrats,
my last 60 elo self-play shrunk to 18 on CCRL, it's the first time I got less than 50%