Stockfish/McBrain 2.5 Release

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

Moderators: hgm, Rebel, chrisw

Dann Corbit
Posts: 12540
Joined: Wed Mar 08, 2006 8:57 pm
Location: Redmond, WA USA

Re: Stockfish/McBrain 2.5 Release

Post by Dann Corbit »

yanquis1972 wrote:hey michael, don't know if the 'readme' on the DL site is new or i just noticed it, but
For overnight analysis, I suggest setting the matefinder option on , Tactical to 3 or 4, and respect to minus 15.
is that for any (generally speaking) position, or extremely difficult positions with a solution?

at what level would tactical be set to to replicate the level of houdini 5 (or 3, since i think this is faster in some positions) tactical search? at what level do you think it goes beyond usefulness in the overwhelming majority of positions?

finally -- what settings would you recommend for, say 1hr analysis? 10min? (i assume default, maybe w/ some level of contempt) 30min? (i don't expect you to have any answer to these, & certainly not a definitive one, just asking out of curiosity)
I guess that all the answers to your questions will depend very much on what sort of machine the program is running.

32 bit with one core?
64 bit with 128 cores?
Something in-between?

I guess it will also vary a lot depending on position. Tactical settings tend to be important for things like zugzwang positions. On the other hand, closed positions usually need immense depth and so tactical settings are counter-productive.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
User avatar
MikeB
Posts: 4889
Joined: Thu Mar 09, 2006 6:34 am
Location: Pen Argyl, Pennsylvania

Re: Stockfish/McBrain 2.5 Release

Post by MikeB »

Dann Corbit wrote:
yanquis1972 wrote:hey michael, don't know if the 'readme' on the DL site is new or i just noticed it, but
For overnight analysis, I suggest setting the matefinder option on , Tactical to 3 or 4, and respect to minus 15.
is that for any (generally speaking) position, or extremely difficult positions with a solution?

at what level would tactical be set to to replicate the level of houdini 5 (or 3, since i think this is faster in some positions) tactical search? at what level do you think it goes beyond usefulness in the overwhelming majority of positions?

finally -- what settings would you recommend for, say 1hr analysis? 10min? (i assume default, maybe w/ some level of contempt) 30min? (i don't expect you to have any answer to these, & certainly not a definitive one, just asking out of curiosity)
I guess that all the answers to your questions will depend very much on what sort of machine the program is running.

32 bit with one core?
64 bit with 128 cores?
Something in-between?

I guess it will also vary a lot depending on position. Tactical settings tend to be important for things like zugzwang positions. On the other hand, closed positions usually need immense depth and so tactical settings are counter-productive.
I'm replying to Dann's post - but this for everyone. Just to pointout, the question being asked is how one find the best move ( and we're not testing engines here - we just want to know the best move)

Even before you do overnight analysis - try these steps - do all of these steps WITHOUT clearing hash in between = you want to benefit from the searches you already did, You also want to do this with big hash - but 4096 is big enough - I have found 8192 almost too big.

1. Default mode - let it run 10 minutes - once it completes - , note the last pv and white moves 2 and 3 for the side you are interested. Moves w1 b1 w2 b2 w3 b3 - note moves w1, w2 and w3. Do searches of these moves will searched by themselves using searchmoves.

1b If you are in a suspected zugzwang position - repeat step 1 with null move off.

1c If you think you have tactical shot involving a sac, repeat step 1 but with matefinder on and tactical set to 3.

1d. Leave the settings as they are and change mutilpv to 3 - - and note the top 3 moves. Test those by themselves using searchmoves

In theory you might now have possibly 12 moves. but in reality you might have 4 or 5 moves or less to real focus on.

Fortress type positions and very deep tactical shots are probably searched best as is.

Shallower tactical shots are often found best by following the procedures above in 1. For sacrifices , use the searchmove option on the sacrifice move you believe might that will work.

Finally - do not be tied to one engine, use K1101, use Houdini 5, etc in addition to SF/McB . They all have their strengths and weaknesses in different posiitons.

Also do not be afraid to walk through moves with the analysis on - moving pieces back and forth and trying different lines WITHOUT clearing HASH - you will be surprised what turns up when you do that.
MikeGL
Posts: 1010
Joined: Thu Sep 01, 2011 2:49 pm

Re: Stockfish/McBrain 2.5 Release

Post by MikeGL »

tpoppins wrote:
MikeGL wrote:both 32-bit labeled old and the generic 32 bit of this version seems to be working fine. Whats the difference between the two?
It's in SF Makefile:

Code: Select all

### 2.2 Architecture specific

<snip>

ifeq ($&#40;ARCH&#41;,x86-32-old&#41;
	arch = i386
endif

ifeq ($&#40;ARCH&#41;,x86-32&#41;
	arch = i386
	prefetch = yes
	sse = yes
endif
Prefetch
SSE
Noted. Thanks for the info.
Vinvin
Posts: 5228
Joined: Thu Mar 09, 2006 9:40 am
Full name: Vincent Lejeune

Re: Stockfish/McBrain 2.5 Release

Post by Vinvin »

MikeB wrote:Details, Source and all flavors of Exe's are available here:

https://github.com/MichaelB7/Stockfish/ ... s/tag/v2.5
Is there a little bug ? when I starts the engine, I got 2 times the TB init message :

Code: Select all

McBrain_2017_v25_x64_modern&#58;
Found 510 tablebases
Found 510 tablebases
User avatar
MikeB
Posts: 4889
Joined: Thu Mar 09, 2006 6:34 am
Location: Pen Argyl, Pennsylvania

Re: Stockfish/McBrain 2.5 Release

Post by MikeB »

Vinvin wrote:
MikeB wrote:Details, Source and all flavors of Exe's are available here:

https://github.com/MichaelB7/Stockfish/ ... s/tag/v2.5
Is there a little bug ? when I starts the engine, I got 2 times the TB init message :

Code: Select all

McBrain_2017_v25_x64_modern&#58;
Found 510 tablebases
Found 510 tablebases
Hmmmm.. doesn't look right. I have it as well. S/b easy to find. Must have fat finger something ...odd.
User avatar
MikeB
Posts: 4889
Joined: Thu Mar 09, 2006 6:34 am
Location: Pen Argyl, Pennsylvania

Re: Stockfish/McBrain 2.5 Release

Post by MikeB »

MikeB wrote:
Vinvin wrote:
MikeB wrote:Details, Source and all flavors of Exe's are available here:

https://github.com/MichaelB7/Stockfish/ ... s/tag/v2.5
Is there a little bug ? when I starts the engine, I got 2 times the TB init message :

Code: Select all

McBrain_2017_v25_x64_modern&#58;
Found 510 tablebases
Found 510 tablebases
Hmmmm.. doesn't look right. I have it as well. S/b easy to find. Must have fat finger something ...odd.
Will be fixed in the next release.
nabildanial
Posts: 126
Joined: Thu Jun 05, 2014 5:29 am
Location: Malaysia

Re: Stockfish/McBrain 2.5 Release

Post by nabildanial »

The time management for this Stockfish with tactical mode on (1~8, not matefinder) is atrocious.
User avatar
MikeB
Posts: 4889
Joined: Thu Mar 09, 2006 6:34 am
Location: Pen Argyl, Pennsylvania

Re: Stockfish/McBrain 2.5 Release

Post by MikeB »

nabildanial wrote:The time management for this Stockfish with tactical mode on (1~8, not matefinder) is atrocious.
Never tested time management for tactical mode since that was mode expected to be used for analysis and not game mode. The tactical mode is driven by multipv code - can you test "official Stockfish"while using multipv set to 8 and see if you have the same observation ? I expect you will, but can you confirm. I'm remote right now and can't test myself. Thank you.
User avatar
Ozymandias
Posts: 1534
Joined: Sun Oct 25, 2009 2:30 am

Re: Stockfish/McBrain 2.5 Release

Post by Ozymandias »

MikeB wrote:Details, Source and all flavors of Exe's are available here:

https://github.com/MichaelB7/Stockfish/ ... s/tag/v2.5
I see you retired the "variety" option, what code would need to be changed, to add this to the main branch?
User avatar
MikeB
Posts: 4889
Joined: Thu Mar 09, 2006 6:34 am
Location: Pen Argyl, Pennsylvania

Re: Stockfish/McBrain 2.5 Release

Post by MikeB »

Ozymandias wrote:
MikeB wrote:Details, Source and all flavors of Exe's are available here:

https://github.com/MichaelB7/Stockfish/ ... s/tag/v2.5
I see you retired the "variety" option, what code would need to be changed, to add this to the main branch?
It was creating a very rare bus error for me so I was not sure if I had it properly code.

I was fudging the value right before the check for next best move:

Code: Select all

	   //Add a little variety to play
      if &#40;variety && value + &#40;variety * 5 * PawnValueEg / 100&#41; >= 0 )
		  value += rand&#40;) % &#40;variety * 5&#41;;
		
	   // Check for a new best move
      if &#40;value > bestValue&#41;
variety was a ucioption with values between 0 and 8