A Stockfish Surprise for the Mac

[MikeB]

Stockfish Surprise for the Mac

Includes:
1. Latest bleeding edge Stockfish-061016

2. Latest bleeding edge Stockfish-061116-01 - with some custom mods

and the Special Surprise...
3. Stockfish DD - Revisited - the Stockfish version dedicated to our friend Don Daily, compiled with the latest optimizations - still a hearty competitor against most competition

Code: Select all

12 Round  - Double Round Robin 
Rank Name                       Elo    +    - games score oppo. draws 
   1 Stockfish DD 64 SSE4.2    3248   92   92    24   75%  3088   50% 
   2 Texel 1.05 64-bit         3225   96   96    24   69%  3090   29% 
   3 Komodo CCT 64-bit         3212   94   94    24   67%  3091   33% 
   4 Hakkapeliitta TCEC v2     3197   96   96    24   65%  3092   29% 
   5 Kōhai 1.0                 3181   93   93    24   63%  3093   33% 
   6 Senpai 1.0                3127   93   93    24   54%  3098   33% 
   7 Stockfish 2.3.1 64bit SS  3107   95   95    24   50%  3099   42% 
   8 Deep HIARCS 14 WCSC       3072   93   93    24   46%  3102   33% 
   9 Crafty-25-1d.0612         3044   98   98    24   42%  3105   25% 
  10 Crafty-25.1.Syzy-0611d    3002  101  101    24   35%  3108   13% 
  11 Crafty-25.1c              2982   99   99    24   31%  3110   29% 
  12 Deep Shredder 12 x64      2955   98   98    24   27%  3112   29% 
  13 Arasan 19.0               2948   98   98    24   27%  3113   29% 

[maxdeg]

Thanks ..

[maxdeg]

Stockfish DD - loaded, runs fine

Stockfish-061016 - "not a valid uci engine ..... "
Stockfish-061116-01 - not a valid uci engine ..... "

MacOS up to date - permissions granted.
Using latest ChessX.
Think my machine has gone bonkers.
Just FYI.

Best wishes .. max

[Eelco de Groot]

Does this include the 1 Elo per keystroke project from the poll that was deleted. I was interested to see the source but it is a private version or files are missing, there are only three things called 'bestand' under Windows from 136 to 140 kB. Okay, no GPL violation then if it does not run

[MikeB]

Stockfish DD - loaded, runs fine

Stockfish-061016 - "not a valid uci engine ..... "
Stockfish-061116-01 - not a valid uci engine ..... "

MacOS up to date - permissions granted.
Using latest ChessX.
Think my machine has gone bonkers.
Just FYI.

Best wishes .. max

Stockfish DD was compiled with the same options as the other two. Sorry , not familiar with ChessX. I would check see if you can start the engine via terminal, type uci and then try "go depth 15". Another idea is to right click it with the mouse and associate the terminal app with it - then try it under the GUI. Every OS update by Apple seems to create new issues and not fix any old issues

[MikeB]

Does this include the 1 Elo per keystroke project from the poll that was deleted. I was interested to see the source but it is a private version or files are missing, there are only three things called 'bestand' under Windows from 136 to 140 kB. Okay, no GPL violation then if it does not run

Yes it does, ELO gain is not holding up at longer tc - but will PM you the source.

[Eelco de Groot]

Thanks Michael, that would be interesting but please if you distribute a modified Stockfish, the GPL requires there is a source somewhere available for everyone.

Maybe an idea to test the changes in Fishtest, if it is just eight characters I think I could manage but I'm sure others would be willing to test your changes.

I was wondering if the HY in the name stood for hyperthreading or maybe Hyatt. I am still searching for that mysterious nullmove improvement from Fruit author Ryan Benitez I think, that was never made public. Chris Conkie wrote about it but never made it public, and Ryan closed his part of Fruit. Maybe it's related to your changes...

[MikeB]

Thanks Michael, that would be interesting but please if you distribute a modified Stockfish, the GPL requires there is a source somewhere available for everyone.

Maybe an idea to test the changes in Fishtest, if it is just eight characters I think I could manage but I'm sure others would be willing to test your changes.

I was wondering if the HY in the name stood for hyperthreading or maybe Hyatt. I am still searching for that mysterious nullmove improvement from Fruit author Ryan Benitez I think, that was never made public. Chris Conkie wrote about it but never made it public, and Ryan closed his part of Fruit. Maybe it's related to your changes...

I was planning to - I considered this a beta version in testing - but your point is well made. I just wanted to test this at longer time control - which is now almost complete results can be found here at the link towards the bottom of the initial post.

http://talkchess.com/forum/viewtopic.php?t=60464


Some good news - the beta version recovered nicely at time 75 sec base , 1.25 sec increment - now shows a 1 point ELO over current base SF. This with testing against all engines - which is hard to verify.

I looked at the fishtest process and for the amount of contributions I expect to make , will not pursue that path. So feel free to to submit on my behalf. Will set up something up later tonight and make the code with the changes freely available. As I indicated earlier - it's about 8 key strokes in just one file.

Will pm you when it's posted.

Edit : misread the chart - show only 1 ELO Point gain - which I found almost impossible to verify. But it's against varied opponents , so maybe it will show in self play ( fishtest).

[MikeB]

source

The 8 key stroke, 8 ELO gain change, line 185 in search.cpp

Code: Select all

  --double r = log(d) * log(mc) / 2;
  ++double r = 35 * log(d) * log(mc-1) / 64 ;

micro tc 10 sec/ .1 sec base - NO ELO CHANGE
tc 75 sec /1.25 sec base - maybe + 8 ELO GAIN ( head to head play )

makefile is different - do make xxxxx at the folder root, not src
folder root should have .d suffix - stockfish-xxx.d
makefile will name the exe 'stockfish-xxx' - misc.cpp version needs to be manually updated with
stockfish.cpp includes all *.cpp files except endgame.cpp

Code: Select all

time control: 6 sec / .1 sec inc
Date: 06/13/16 : 22:26:56
Rank Name                  Elo    +    - games score oppo. draws 
   1 Stockfish 06-10-16   3100    4    4  3400   50%  3100   57% 
   2 Stockfish 061116-01  3100    4    4  3400   50%  3100   57% 

[Eelco de Groot]

Thanks for your source and submitting it to GitHub Michael.
I just copied your search.cpp change and submitted a test on Fish Framework. Hope it will run!

Regards, Eelco