Stockfish at the top

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

Ramiras

Stockfish at the top

Post by Ramiras »

User avatar
AdminX
Posts: 6340
Joined: Mon Mar 13, 2006 2:34 pm
Location: Acworth, GA

Re: Stockfish at the top

Post by AdminX »

Well a Huge Congrats to Stockfish !!!! :D
"Good decisions come from experience, and experience comes from bad decisions."
__________________________________________________________________
Ted Summers
IanO
Posts: 496
Joined: Wed Mar 08, 2006 9:45 pm
Location: Portland, OR

Re: Stockfish at the top

Post by IanO »

Ramiras wrote:http://www.computerchess.org.uk/ccrl/40 ... t_all.html
Why is F***b*** not mentioned?
No reputable rating list currently accepts The Engines Which Must Not Be Named, due to the likelyhood that their lineage is tainted with code stolen from our dark lord and master, Rybka. Do not speak their names or you may be stolen away in the dead of night! :evil:
muxecoid
Posts: 150
Joined: Sat Jan 30, 2010 10:54 am
Location: Israel

Re: Stockfish at the top

Post by muxecoid »

Ramiras wrote:http://www.computerchess.org.uk/ccrl/40 ... t_all.html
Why is F.... not mentioned?
Great news.

The engines that should not be named are very strong at short time controls while losing strength at long time control, I guess in 40/40 Stockfish is unconditionally the best.

Some guys on other forum (forum that should not be linked to) say that if you play with SF1.7 parameters to increase aggressiveness it beats forbidden engines even with short time controls.
Ramiras

Re: Stockfish at the top

Post by Ramiras »

Thanks! Any suggestions on what values to use for aggressiveness - cowardness - no of ht's etc as an experiment to beat the unnameable clone ? :roll:
Ralph Stoesser
Posts: 408
Joined: Sat Mar 06, 2010 9:28 am

Re: Stockfish at the top

Post by Ralph Stoesser »

Congrats to the Stockfish team!

In next version please add UCI param killAllClonesRegardlessOfTimeControl.
Default should be set to true. 8-)
Terry McCracken
Posts: 16465
Joined: Wed Aug 01, 2007 4:16 am
Location: Canada

Re: Stockfish at the top

Post by Terry McCracken »

IanO wrote:
Ramiras wrote:http://www.computerchess.org.uk/ccrl/40 ... t_all.html
Why is F***b*** not mentioned?
No reputable rating list currently accepts The Engines Which Must Not Be Named, due to the likelyhood that their lineage is tainted with code stolen from our dark lord and master, Rybka. Do not speak their names or you may be stolen away in the dead of night! :evil:
If only that were true... :lol:
Terry McCracken
aturri
Posts: 85
Joined: Wed Dec 30, 2009 11:35 pm

Re: Stockfish at the top

Post by aturri »

Ramiras wrote:http://www.computerchess.org.uk/ccrl/40 ... t_all.html
Why is Firebird not mentioned?
Well, I am a bit surprised with that list. What puzzles me is the rare speed-up shown by Stockfish 1.7.1 which no other engine shows, specially the previous versions of SF (like 1.5.1 and 1.6.3)

I have built the following table based on the present values shown on the 'CCRL 40/40' list. You see the best engines and versions that have rating for both 1CPU, 2CPU and 4CPU, to calculate the (sort of) speed-up when running 2CPU and 4CPU. The last column is the relative increment of speed-up from 2CPU to 4CPU compared to the 1CPU to 4CPU one.

Code: Select all

---------------------------+--------+--------+--------+---------+---------+-----------+
          Engine           |  1CPU  |  2CPU  |  4CPU  |2CPU-1CPU|4CPU-2CPU|(4-2)/(4-1)|
---------------------------+--------+--------+--------+---------+---------+-----------+
Stockfish 1.7.1 64-bit     |  3080  |  3110  |  3249  |    30   |   139   |   82,2%   |
Stockfish 1.6.3 64-bit     |  3074  |  3111  |  3145  |    37   |    34   |   47,9%   |
Stockfish 1.5.1 64-bit     |  2996  |  3051  |  3078  |    55   |    27   |   32,9%   |
---------------------------+--------+--------+--------+---------+---------+-----------+
Rybka 3 64-bit             |  3153  |  3179  |  3232  |    26   |    53   |   67,1%   |
Rybka 2.3.2a 64-bit        |  3068  |  3089  |  3128  |    21   |    39   |   65,0%   |
Rybka 2.2n2 64-bit         |  3010  |  3056  |  3083  |    46   |    27   |   37,0%   |
---------------------------+--------+--------+--------+---------+---------+-----------+
Naum 4.2 64-bit            |  3102  |  3080  |  3174  |   -22   |    94   |  130,6%   |
Naum 4 64-bit              |  3059  |  3108  |  3153  |    49   |    45   |   47,9%   |
Naum 3.1 64-bit            |  2966  |  3022  |  3076  |    56   |    54   |   49,1%   |
Naum 3 64-bit              |  2962  |  3027  |  3066  |    65   |    39   |   37,5%   |
Naum 2.2 64-bit            |  2904  |  2975  |  3008  |    71   |    33   |   31,7%   |
---------------------------+--------+--------+--------+---------+---------+-----------+
Zappa Mexico II 64-bit     |  2966  |  3017  |  3074  |    51   |    57   |   52,8%   |
Zappa Mexico 64-bit        |  2941  |  3017  |  3067  |    76   |    50   |   39,7%   |
Zap!Chess Zanzibar 64-bit  |  2905  |  3025  |  3050  |   120   |    25   |   17,2%   |
---------------------------+--------+--------+--------+---------+---------+-----------+
Thinker 5.4D Inert 64-bit  |  2961  |  2970  |  3034  |     9   |    64   |   87,7%   |
Thinker 5.4C Inert 64-bit  |  2973  |  3004  |  3036  |    31   |    32   |   50,8%   |
Thinker 5.4a Inert 64-bit  |  2959  |  3006  |  3012  |    47   |     6   |   11,3%   |
---------------------------+--------+--------+--------+---------+---------+-----------+
Spark 0.3 64-bit           |  2945  |  2961  |  3035  |    16   |    74   |   82,2%   |
---------------------------+--------+--------+--------+---------+---------+-----------+
Deep Sjeng WC2008 64-bit   |  2943  |  3046  |  3031  |   103   |   -15   |  -17,0%   |
---------------------------+--------+--------+--------+---------+---------+-----------+
Deep Shredder 11 64-bit    |  2947  |  2994  |  3024  |    47   |    30   |   39,0%   |
---------------------------+--------+--------+--------+---------+---------+-----------+
Toga II 1.4.1SE            |  2933  |  2963  |  3010  |    30   |    47   |   61,0%   |
Toga II 1.4 beta5c         |  2909  |  2969  |  2992  |    60   |    23   |   27,7%   |
---------------------------+--------+--------+--------+---------+---------+-----------+
Hiarcs 12                  |  2914  |  2987  |  3004  |    73   |    17   |   18,9%   |
Hiarcs 11.1                |  2894  |  2942  |  2983  |    48   |    41   |   46,1%   |
---------------------------+--------+--------+--------+---------+---------+-----------+
Glaurung 2.2 64-bit        |  2889  |  2972  |  3004  |    83   |    32   |   27,8%   |
---------------------------+--------+--------+--------+---------+---------+-----------+
Grapefruit 1.0 32-bit      |  2934  |  2973  |  2962  |    39   |   -11   |  -39,3%   |
---------------------------+--------+--------+--------+---------+---------+-----------+
(Deep) Fritz 10            |  2885  |  2953  |  2934  |    68   |   -19   |  -38,8%   |
---------------------------+--------+--------+--------+---------+---------+-----------+
Deep Junior 10             |  2911  |  2901  |  2913  |   -10   |    12   |  600,0%   |
---------------------------+--------+--------+--------+---------+---------+-----------+
The speed-up of Stockfish 1.7.1 is anormally high compared with all the other engines, both in absolute increment (¡139! when duplicating cores) and relative increment (>80%)

[Note that some relative increments have weird values due to some rare results, e.g., Deep Fritz 10, Deep Sjeng WC2008 and Grapefruit with a negative speed-up when scaling from 2CPU to 4CPU, or as in the case of Naum 4.2 due to an abnormally low result for the 2CPU executions)]

I would like confirmation of the values, as for example, if the 'Stockfish 1.7.1 64-bit' rating was 3149 instead of 3249 (just one digit) the result would be much more concordant with the rest of speed-ups.

If no typo is confirmed, then I cannot imaging what could happen when Stockfish 1.7.1 64-bit will run on a 8 or even 16 core computer!!! :)

[Edited a typo on the table)
Uri Blass
Posts: 10282
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Stockfish at the top

Post by Uri Blass »

I think that you simply ignore the number of games

The real difference between 4 cpu and 1 cpu is clearly smaller for stockfish based on more games at 40/4 time control

http://www.computerchess.org.uk/ccrl/40 ... +opponents

Stockfish 1.7.1 64-bit 4CPU 3229 +17 −16 70.2% −140.3 36.0% 1303
Stockfish 1.7.1 64-bit 3136 +21 −20 73.9% −167.5 35.4% 878


The rating of stockfish at 40/40(espacially the 1 cpu and 2 cpu that do not have even 100 games is not reliable)

Here is the statistics for stockfish 1 cpu and 2 cpu and 4 cpu at 40/40
I expect the rating of stockfish 1 cpu and 2 cpu to improve with more games

Stockfish 1.7.1 64-bit 4CPU 3249 +37 −36 71.0% −136.8 43.5% 248
Stockfish 1.7.1 64-bit 2CPU 3110 +111 −111 50.0% +15.3 54.5% 22
Stockfish 1.7.1 64-bit 3080 +66 −65 58.8% −55.1 50.0% 68
muxecoid
Posts: 150
Joined: Sat Jan 30, 2010 10:54 am
Location: Israel

Re: Stockfish at the top

Post by muxecoid »

Looks like the quality of play improves with time faster, than for other engines. I wonder how good human-free Stockfish 1.7.1 is at correspondence chess.

I'd like to see CCRL for 40/40 Intel i940 and not 40/40 Atlon X2 4600+.