Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Computer Chess / Stockfish 1.6.3 JA
- - By Morpheus (****) Date 2010-02-03 21:53
Hi,

What's the new in Stockfish 1.6.3 JA , What diferences between Stockfish 1.6.3 JA and old version Stockfish 1.6.2 JA ??
Somebody know? And now....how aboute ELO ? how much is now?

  1 Rybka 3 Dyn,Rybka 3.ctg        3201   28   27   565   79%  2971   27%
   2 Rybka 3,Rybka3.ctg             3188   28   27   557   79%  2965   30%
   3 Rybka 3 C0,Rybka 3.ctg         3172   28   27   438   73%  3022   41%
  4 Stockfish 1.6.2,Salvo 1.6.bin  3161   25   24   538   66%  3056   43%

and new verision ? maybe Sedat know?

Sekos
Parent - - By M ANSARI (*****) Date 2010-02-03 22:40
I tested on 8 cores and it was around +30 ELO.  I don't think you will see much improvement except for 8 core systems, although I could be wrong.  In task manager Stockfish 1.6 is not at 100% while 1.6.3 is, this would make it seem like the latest version has better MP scaling.
Parent - By Morpheus (****) Date 2010-02-03 22:46
thnax M ANSARI
"In task manager Stockfish 1.6 is not at 100% while 1.6.3 is" I check this earlier and I must confirm that,

I agreee with YOU, so can You give link for yours game at this engine

Sekos
Parent - By NATIONAL12 (Gold) Date 2010-02-03 22:59
agree i have posted this,also on TalkChess.
Parent - - By Dave Felton (*) Date 2010-02-04 07:43 Edited 2010-02-04 07:51
The author said that this is a bugfix release only (1.63).  The Elo difference should be zero when compared to 1.62.  Any small Elo differences you see might be due to errors from using a small sample size.

Authors comments on Jim's page:

What's New:
Maintenance release.
The aim of this release is to fix all the reported bugs that have accumulated.
It is 100% functionally and speed equivalent to 1.6.2 so there is NO ELO DIFFERENCE.

http://homepages.tesco.net/henry.ablett/jims.html
Parent - - By Morpheus (****) Date 2010-02-04 09:48
Good POINT, thanx for Response Dave .

Sekos
Parent - By Dave Felton (*) Date 2010-02-04 13:34
Your welcome mate.

To be honest I was a bit disappointed to see there was no difference myself. 

I am really looking forward to next versions of this and Komodo.

Dave

PS  And of course Rybka 4, but I figure that could be a while ;)
Parent - - By M ANSARI (*****) Date 2010-02-04 16:44
With 8 cores, there is a difference in ELO.  I am not sure how much this ELO gain is but probably around 20 to 30 ELO plus.  The reason is quite simple, only the latest Stockfish release enables all 8 CPU's.  In previous versions it only enables 7 CPU's.  You can see that in Task manager.  In my testing it was +28 ELO stronger, but my tests were 3_1 so that is pretty fast time control.  Most likely at longer time controls it will be less and I would guess 1 extra usable CPU to worth about 15 ELO to 20 ELO.  With 4 cores I think there should be very little difference if any.
Parent - - By Prima (****) Date 2010-02-04 18:29
Good for you Maj ANSARI :). By the way, just in case no one knew about the "other" StockFish 1.6.3s by Dann Corbit, here it is from the TalkChess threads http://www.talkchess.com/forum/viewtopic.php?t=32316

Also uploaded here
Parent - - By RFK (Gold) Date 2010-02-05 00:58
Hi Prima,

I recognize the overwhelming use of 64 bit- but just for those of us who are still lingering and waiting to upgrade to 64 bit  here is also a link for Dan's Stockfish 1.6.3.s with a 32bit engine included.

http://www.talkchess.com/forum/viewtopic.php?p=328088&highlight=#328088
Parent - By Prima (****) Date 2010-02-05 05:04

> I recognize the overwhelming use of 64 bit- but just for those of us who are still lingering and waiting to upgrade to 64 bit  here is also a link for Dan's Stockfish 1.6.3.s with a 32bit engine included.


Thanks sidserious.

I wasn't aware the 32-bit StockFish 1.6.3s counterpart was compiled. I appreciate this and will keep this 32-bit SF-SS for keeps-sake :) I have 64-bit OS Quad PC though.
Parent - By Banned for Life (Gold) Date 2010-02-05 00:15
15 to 20 Elo from a speedup of 7 to 8 cores alone, i.e. less than 15 percent, seems highly unlikely. The gain would have to be partially due to poor utilization of cores 5, 6, and 7 in 1.6.2.
Parent - - By FWCC (***) Date 2010-02-04 22:04
There should be some increase in strength.There is a small reported increase due to these bug fixes.
Parent - By Prima (****) Date 2010-02-05 05:14

> There should be some increase in strength.There is a small reported increase due to these bug fixes.


Well, according to the SF Team, the releases from 1.6 to 1.6.3 have all been maintenance release - with the exception of Dann Corbit's SF version of course. Still, most people have reported an ELO increase in spite of SF 1.6+ being maintenance. Maybe there's ELO gains, maybe not. I haven't tested all SF-1.6+ versions to verify these reports.
- - By Kreuzfahrtschiff (***) Date 2010-02-04 20:13
i view such engines, (esp. Stockfish 1.6.3s(dc)) from correspondence players look, and its amazing. it found nearly every difficult move after 10-30 min, where rybka needs half a day or more  :)
Parent - - By Prima (****) Date 2010-02-05 00:24
Apparently the SF-Smooth Scaling implementation from Dann Corbit has some positive impacts. Hopefully smooth scaling will find its way into the main line SF sooner.
Parent - - By M ANSARI (*****) Date 2010-02-05 03:44
To be honest I still don't know how SS affects the engine.  It seemed to test really well last time I tested it, but others have claimed that it is not much ELO gain.  I think they are planning to implement it after they figure out which is the best setup.  So far the last few SF upgrades have been minor cosmetic and bug fix upgrades, but I am sure they are planning a "real" strength upgrade very soon.  That should pull in some good ELO gains.
Parent - By Prima (****) Date 2010-02-05 05:08

> but I am sure they are planning a "real" strength upgrade very soon.  That should pull in some good ELO gains.


Same hopes here for SF. I can hardly wait for the next major SF release rather than the maintenance released versions. My main concern with SF is the evaluations that most seem unsatisfied with. I hope the SF Teams considers too - if possible.
Parent - - By Master Om (Bronze) Date 2010-02-05 04:02
As you said from Corrrespondence point of View how do you use it ?. In IDeA or Infinite Analysis ?
Parent - By Kreuzfahrtschiff (***) Date 2010-02-05 11:20
Infinite Analysis and i move forward myself, and was analysing known positions, where rybka computed 3 weeks :) and found better moves after 2 weeks
- - By Martin Thoresen (***) Date 2010-02-05 04:48
From CCRL 40/4 list. I don't see how the SS version is any stronger. It has 8 more elo points, but also 8 elo weaker opposition overall.

Name                                 Rating ELO  +  −  Score  Average Opponent  Draws  Games
Stockfish 1.6s JA 64-bit 4CPU  3187  +14  −14  68.1%        −126.7           37.3%  1793
Stockfish 1.6.2 64-bit 4CPU     3179  +14  −14  66.7%        −118.3           36.0%  1795
Parent - By Uly (Gold) Date 2010-02-05 06:12

> It has 8 more elo points, but also 8 elo weaker opposition overall.


That's not important. However, the +14 -14 really indicate that there's no proof of the SS version is stronger, but there's a 78.8% chance that it indeed is stronger.
Up Topic The Rybka Lounge / Computer Chess / Stockfish 1.6.3 JA

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill