Stockfish Tuning

on Mon Dec 09, 2013 8:17 pm by · on Mon Dec 09, 2013 8:17 pm by

@MALIBU wrote:COWARDICE value.

The default value (100).

on Mon Dec 09, 2013 8:20 pm by · on Mon Dec 09, 2013 8:20 pm by

@Bob wrote:
@MALIBU wrote:COWARDICE value.
The default value (100).

on Mon Dec 09, 2013 10:51 pm by · on Mon Dec 09, 2013 10:51 pm by

Aggressiveness = 90

Stockfish is coming at ya, Houdini!

on Tue Dec 10, 2013 6:30 am by · on Tue Dec 10, 2013 6:30 am by

VocalTechnique wrote:Aggressiveness = 90

Stockfish is coming at ya, Houdini!

Sorry, but I don't agree with that setting. It may work for short Time control games, but not for Long-Time Cotrol games.
It just gives more draws ! Default is best, like it or not !

on Tue Dec 10, 2013 9:34 am by · on Tue Dec 10, 2013 9:34 am by

I always use the default settings. There are lots of testers around the world helping the SF development. They have different kinds of setup in their computers and play many different time controls from short to long. These results are permanently updating the main database. For example if they are doing tests using agg=90 and the results are good the authors of the engine change the internal value of agg to represent the optimal 90 value, but you always will see 100 as a default in the engine configure mask. So there is no point in changing any values of the default. They are already doing this for us!!! cool

http://tests.stockfishchess.org/tests

on Tue Dec 10, 2013 9:39 am by · on Tue Dec 10, 2013 9:39 am by

@Bob wrote:
Code:
Games Completed = 500 of 500 (Avg game length = 31.066 sec) Settings = RR/32MB/4000ms+200ms/M 700000cp for 7 moves, D 150 moves/EPD:D:\LittleBlitzer\openings.epd(4000) Time = 4020 sec elapsed, 0 sec remaining 1. Stockfish 091213 64 (A90) 265.5/500 127-96-277 (L: m=96 t=0 i=0 a=0) (D: r=220 i=37 f=10 s=0 a=10) (tpm=244.0 d=18.91 nps=1190073) 2. Stockfish 091213 64 234.5/500 96-127-277 (L: m=127 t=0 i=0 a=0) (D: r=220 i=37 f=10 s=0 a=10) (tpm=244.7 d=18.84 nps=1190348)

Stockfish 091213 64 (A90); Aggressiveness = 90 (Ahmed's suggestion).

Great!!!

Stockfish 10-12-2013 64bit vs. Stockfish-A90 10-12-2013 64bit Match is underway
Stockfish-A90 10-12-2013 64bit: Aggressiveness = 90
Time Control: 3 Minutes
1000 Games.

on Tue Dec 10, 2013 11:53 am by · on Tue Dec 10, 2013 11:53 am by

Thx Vladimir.

on Tue Dec 10, 2013 4:59 pm by · on Tue Dec 10, 2013 4:59 pm by

Thank you Vladimir

on Tue Dec 10, 2013 5:01 pm by · on Tue Dec 10, 2013 5:01 pm by

@MadMax777 wrote:I always use the default settings. There are lots of testers around the world helping the SF development. They have different kinds of setup in their computers and play many different time controls from short to long. These results are permanently updating the main database. For example if they are doing tests using agg=90 and the results are good the authors of the engine change the internal value of agg to represent the optimal 90 value, but you always will see 100 as a default in the engine configure mask. So there is no point in changing any values of the default. They are already doing this for us!!!

http://tests.stockfishchess.org/tests

Good explanation!

on Tue Dec 10, 2013 8:09 pm by · on Tue Dec 10, 2013 8:09 pm by

@Vladimir wrote:
@Bob wrote:
Code:
Games Completed = 500 of 500 (Avg game length = 31.066 sec) Settings = RR/32MB/4000ms+200ms/M 700000cp for 7 moves, D 150 moves/EPD:D:\LittleBlitzer\openings.epd(4000) Time = 4020 sec elapsed, 0 sec remaining 1. Stockfish 091213 64 (A90) 265.5/500 127-96-277 (L: m=96 t=0 i=0 a=0) (D: r=220 i=37 f=10 s=0 a=10) (tpm=244.0 d=18.91 nps=1190073) 2. Stockfish 091213 64 234.5/500 96-127-277 (L: m=127 t=0 i=0 a=0) (D: r=220 i=37 f=10 s=0 a=10) (tpm=244.7 d=18.84 nps=1190348)

Stockfish 091213 64 (A90); Aggressiveness = 90 (Ahmed's suggestion).

Great!!!

Stockfish 10-12-2013 64bit vs. Stockfish-A90 10-12-2013 64bit Match is underway
Stockfish-A90 10-12-2013 64bit: Aggressiveness = 90
Time Control: 3 Minutes
1000 Games.

Eagerly waiting for final result.. thumb up

on Tue Dec 10, 2013 8:29 pm by · on Tue Dec 10, 2013 8:29 pm by

my test:

Quad i7
Time control 5min + 3 sec.
Fritz 14 GUI

Code:: Stockfish DD x64 (Aggressiveness 90) - Stockfish DD x64 23.5 - 16.5 +12/=23/-5 58.75%

on Tue Dec 10, 2013 9:49 pm by · on Tue Dec 10, 2013 9:49 pm by

Very nice test, Waelsae. Thank you!

on Tue Dec 10, 2013 11:19 pm by · on Tue Dec 10, 2013 11:19 pm by

@dass wrote:
@Vladimir wrote:
@Bob wrote:
Code:
Games Completed = 500 of 500 (Avg game length = 31.066 sec) Settings = RR/32MB/4000ms+200ms/M 700000cp for 7 moves, D 150 moves/EPD:D:\LittleBlitzer\openings.epd(4000) Time = 4020 sec elapsed, 0 sec remaining 1. Stockfish 091213 64 (A90) 265.5/500 127-96-277 (L: m=96 t=0 i=0 a=0) (D: r=220 i=37 f=10 s=0 a=10) (tpm=244.0 d=18.91 nps=1190073) 2. Stockfish 091213 64 234.5/500 96-127-277 (L: m=127 t=0 i=0 a=0) (D: r=220 i=37 f=10 s=0 a=10) (tpm=244.7 d=18.84 nps=1190348)

Stockfish 091213 64 (A90); Aggressiveness = 90 (Ahmed's suggestion).

Great!!!

Stockfish 10-12-2013 64bit vs. Stockfish-A90 10-12-2013 64bit Match is underway
Stockfish-A90 10-12-2013 64bit: Aggressiveness = 90
Time Control: 3 Minutes
1000 Games.

Eagerly waiting for final result..

+1

on Wed Dec 11, 2013 9:08 am by · on Wed Dec 11, 2013 9:08 am by

Stockfish-A90 10-12-2013 64bit vs. Stockfish 10-12-2013 64bit
555 out of 1000 games played
http://www.chess2u.com/t7901-stockfish-a90-10-12-2013-64bit-vs-stockfish-10-12-2013-64bit#49279

on Thu Dec 12, 2013 10:21 am by · on Thu Dec 12, 2013 10:21 am by

Stockfish-A90 10-12-2013 64bit vs. Stockfish 10-12-2013 64bit
1000 out of 1000 games played
http://www.chess2u.com/t7901-stockfish-a90-10-12-2013-64bit-vs-stockfish-10-12-2013-64bit#49305

on Thu Dec 12, 2013 8:16 pm by · on Thu Dec 12, 2013 8:16 pm by

Both your 3 min tests and Bobs tests are the best. At least 1000 games make your tests stand in my mind. Thank you for taking your time and resources for us.

on Fri Dec 13, 2013 12:21 pm by · on Fri Dec 13, 2013 12:21 pm by

my test:

Quad i7
Time control 5min + 3 sec.
Fritz 14 GUI

Code:: Stockfish DD x64 (Aggressiveness 90) - Stockfish DD x64 80.5 - 68.5 +32/=97/-20 54.03%

on Wed Apr 30, 2014 2:10 am by · on Wed Apr 30, 2014 2:10 am by

Can anyone explain to me what the different settings mean?

Contempt
Cowardice
Aggressiveness
Pawn structure
Space
Min split depth

etc....

and what the best numbers are

Thanks thumb up

on Wed Apr 30, 2014 2:29 am by · on Wed Apr 30, 2014 2:29 am by

Engine Parameters

This is a list of Stockfish engine parameters and their explanations.

http://support.stockfishchess.org

Write Debug Log Boolean, Default: False
Whether or not to write a debug log.

Write Search Log Boolean, Default: False
Whether or not to write a search log.

Search Log Filename String, Default: SearchLog.txt
The file name of the search log.

Book File String, Default: book.bin
The file name of the opening book file.

Best Book Move Boolean, Default: False
Whether or not to always play the best move from the opening book. False will lead to more variety in opening play.

Contempt Factor Integer, Default: 0, Min: -50, Max: 50
Roughly equivalent to "optimism." Positive values of contempt favor more "risky" play, while negative values will favor draws. Zero is neutral.

Mobility (Midgame) Integer, Default: 100, Min: 0, Max: 200

Mobility (Endgame) Integer, Default: 100, Min: 0, Max: 200

Pawn Structure (Midgame) Integer, Default: 100, Min: 0, Max: 200

Pawn Structure (Endgame) Integer, Default: 100, Min: 0, Max: 200

Passed Pawns (Midgame) Integer, Default: 100, Min: 0, Max: 200

Passed Pawns (Endgame) Integer, Default: 100, Min: 0, Max: 200

Space Integer, Default: 100, Min: 0, Max: 200

Aggressiveness Integer, Default: 100, Min: 0, Max: 200
Degree of agressiveness.

Cowardice Integer, Default: 100, Min: 0, Max: 200
Degree of cowardice.

Min Split Depth Integer, Default: 0, Min: 0, Max: 12

Max Threads per Split Point Integer, Default: 5, Min: 4, Max: 8

Threads Integer, Default: 1, Min: 1, Max: 64
The number of threads to use during the search. This number should be set to the number of cores in your CPU.

Idle Threads Sleep Boolean, Default: True for Dev versions, False for Release versions
If this is set to true, threads are suspended when there is no work to do. This saves CPU power consumption, but waking a thread takes a small bit of time. For maximum performance, set this option to false, but if you need to reduce power consumption (i.e. on mobile devices) set this option to true.

Hash Integer, Default: 32, Min: 1, Max: 8192
The amount of memory to use for the hash during search, specified in MB (megabytes). This number should be smaller than the amount of physical memory for your system.

Ponder Boolean, Default: True
Whether or not the engine should analyze when it is the opponent's turn.

OwnBook Boolean, Default: False
Whether or not the engine should use the opening book. Also see the Book File option.

MultiPV Integer, Default: 1, Min: 1, Max: 500
The number of alternate lines of analysis to display. Specify 1 to just get the best line. Asking for more lines slows down the search.

Skill Level Integer, Default: 20, Min: 0, Max: 20
How well you want Stockfish to play. At level 0, Stockfish will make dumb moves. Level 20 is best/strongest play.

Emergency Move Horizon Integer, Default: 40, Min: 0, Max: 50

Emergency Base Time Integer, Default: 60, Min: 0, Max: 30000

Emergency Move Time Integer, Default: 30, Min: 0, Max: 5000

Minimum Thinking Time Integer, Default: 20, Min: 0, Max: 5000
The minimum amount of time to analyze, in milliseconds.

Slow Mover Integer, Default: 70, Min: 10, Max: 1000

UCI_Chess960 Boolean, Default: False
Whether or not Stockfish should play using Chess 960 mode. Chess 960 is a chess variant where the back ranks are scrambled.

on Wed Apr 30, 2014 2:32 am by · on Wed Apr 30, 2014 2:32 am by

Ok thank you :)

So what effect would setting the pawn structure and space parameters lower, to around 50 have?

Thanks.

on Wed Apr 30, 2014 2:35 am by · on Wed Apr 30, 2014 2:35 am by

Ahmed Kamal : Set Aggressiveness to 90 from 100.

on Wed Apr 30, 2014 2:38 am by · on Wed Apr 30, 2014 2:38 am by

Ok thank you.

I tested it with contempt = 50
cowardice = 0
and the rest were all default setting and it seemed to improve,beating the default version 60% of the time