New 6-piece tablebases

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

syzygy
Posts: 5647
Joined: Tue Feb 28, 2012 11:56 pm

Re: New 6-piece tablebases

Post by syzygy »

syzygy wrote:A few things need to be fixed. I will try to do that today.
I fixed a few things and updated to the latest Stockfish sources.

Missing DTZ tables could actually cause an exception. This should be fixed now.

If DTZ tables are missing, once a tablebase position has been reached Stockfish will now try to make progress by searching for a won position with one piece less (using the WDL tables). This should normally be sufficient to convert a won position into a mate, but in exceptional cases Stockfish might not understand the position and play a drawing/losing move when the position is won/drawn.

When I have time I will make this fallback more sophisticated by additionally using the WDL tables to filter out "bad" moves at the root. This will prevent Stockfish from playing a drawing/losing move when the position is won/drawn, but it could still happen that a win is too complicated and the position is unnecessarily drawn by the 50-move rule. With DTZ tables this cannot happen, as these tables ensure progress.
Sharaf_DG
Posts: 30
Joined: Fri Aug 02, 2013 11:03 am

Re: New 6-piece tablebases

Post by Sharaf_DG »

Some results with latest update of last night vs previous update of Friday 20/09/13
1. RTB-130920 1963-2040-5997
2. RTB-130922 2040-1963-5997 [/color] [/color][/color]
ELO=+3....Improvement noted :D
IQ
Posts: 162
Joined: Thu Dec 17, 2009 10:46 am

Re: New 6-piece tablebases

Post by IQ »

Interesting would be a test of 6 piece Syzygy Stockfish 4 against plain Stockfish 4. Did you run such a test?
Sharaf_DG wrote:Some results with latest update of last night vs previous update of Friday 20/09/13
1. RTB-130920 1963-2040-5997
2. RTB-130922 2040-1963-5997 [/color] [/color][/color]
ELO=+3....Improvement noted :D
Sharaf_DG
Posts: 30
Joined: Fri Aug 02, 2013 11:03 am

Re: New 6-piece tablebases

Post by Sharaf_DG »

IQ wrote:Interesting would be a test of 6 piece Syzygy Stockfish 4 against plain Stockfish 4. Did you run such a test?
Test running, will post soon
phenri
Posts: 284
Joined: Tue Aug 13, 2013 9:44 am

Re: New 6-piece tablebases

Post by phenri »

Thanks Ronald

Here's a new compilation based on the latest sotckifish dated 23 September 2013. Source code included

Stockfish-master_130923_sbase_support.7z
phenri
Posts: 284
Joined: Tue Aug 13, 2013 9:44 am

Re: New 6-piece tablebases

Post by phenri »

I'm anxious to see added new options listed in the ReadMe :)
User avatar
Marek Soszynski
Posts: 582
Joined: Wed May 10, 2006 7:28 pm
Location: Birmingham, England

Re: New 6-piece tablebases

Post by Marek Soszynski »

phenri wrote:Thanks Ronald

Here's a new compilation based on the latest sotckifish dated 23 September 2013. Source code included

Stockfish-master_130923_sbase_support.7z
Thanks to everyone involved!

The sources compile in Linux, and the Syzygy tablebases are accessed okay.
Marek Soszynski
Vinvin
Posts: 5239
Joined: Thu Mar 09, 2006 9:40 am
Full name: Vincent Lejeune

Re: New 6-piece tablebases

Post by Vinvin »

phenri wrote:Thanks Ronald

Here's a new compilation based on the latest sotckifish dated 23 September 2013. Source code included

Stockfish-master_130923_sbase_support.7z
I downloaded Sbase from here : http://olympuschess.com/egtb/sbases/sbases345.tar.bz2
I set the sbase path in the parameters but this endgame gives :
[d]8/4P3/8/1bk1K2B/8/8/8/8 w - - 0 69

Code: Select all

Analysis by Stockfish 230913 64:

69.Re6 
  +-  (3.83)   Profondeur: 1/2   00:00:00
69.Re6 Fc4+ 70.Rf6 Fb5 71.Re5 Fc6 
  +-  (3.75)   Profondeur: 5/6   00:00:00
69.Re6 Fc4+ 70.Rf6 Fb5 71.Re5 Fc6 
  +-  (3.75)   Profondeur: 5/6   00:00:00
69.Re6 Fc4+ 70.Rf6 Fb5 71.Re5 Fc6 
  +-  (3.75)   Profondeur: 5/6   00:00:00
69.Re6 Fc4+ 70.Rf6 Fb5 71.Re5 Fc6 
  +-  (3.75)   Profondeur: 5/6   00:00:00
69.Re6 Fc4+ 70.Rf6 Fb5 71.Re5 Fd7 72.Fg6 Rc6 
  +-  (3.67)   Profondeur: 7/8   00:00:00  2kN, tb=1
69.Re6 Fc4+ 70.Rf6 Fb5 71.Re5 Fd7 72.Fg6 Rc6 
  +-  (3.67)   Profondeur: 7/8   00:00:00  2kN, tb=1
69.Ff7 Fd7 70.Fd5 Fb5 71.Re6 Fe8 72.Fe4 Rd4 73.Ff3 Rc5 74.Fd5 Fg6 75.Fe4 Fe8 76.Re5 Fb5 77.Fd5 Fe8 78.Re6 
  +-  (3.47)   Profondeur: 17/19   00:00:00  39kN, tb=1
69.Ff7 Fd7 70.Fd5 Fb5 71.Re6 Fe8 72.Fe4 Rd4 73.Ff3 Rc5 74.Fd5 Fg6 75.Fe4 Fe8 76.Re5 Fb5 77.Fd5 Fe8 78.Re6 
  +-  (3.47)   Profondeur: 17/19   00:00:00  39kN, tb=1
69.Ff7 Fd7 70.Fd5 Fb5 71.Re6 Fe8 72.Fe4 Rd4 73.Ff3 Rc5 74.Fd5 Fg6 75.Fe4 Fe8 76.Re5 Fb5 77.Fd5 Fe8 78.Re6 
  +-  (3.47)   Profondeur: 17/19   00:00:00  39kN, tb=1
69.Ff7 Fd7 70.Fd5 Fb5 71.Re6 Fe8 72.Fe4 Rd4 73.Ff3 Rc5 74.Fd5 Fg6 75.Fe4 Fe8 76.Re5 Fb5 77.Fd5 Fe8 78.Re6 
  +-  (3.47)   Profondeur: 17/19   00:00:00  39kN, tb=1
69.Ff7 Fd7 70.Fd5 Fb5 71.Re6 Fe8 72.Fe4 Rd4 73.Ff3 Rc5 74.Fd5 Fg6 75.Fe4 Fe8 76.Re5 Fb5 77.Fd5 Fe8 78.Re6 
  +-  (3.47)   Profondeur: 17/19   00:00:00  39kN, tb=1
69.Ff7 Fd7 70.Fd5 Fb5 71.Re6 Fe8 72.Fe4 Rd4 73.Ff3 Rc5 74.Fd5 Fg6 75.Fe4 Fe8 76.Re5 Fb5 77.Fd5 Fe8 78.Re6 
  +-  (3.47)   Profondeur: 17/19   00:00:00  39kN, tb=1
69.Ff7 Fd7 70.Fd5 Fb5 71.Re6 Fe8 72.Fe4 Rd4 73.Ff3 Rc5 74.Fd5 Fg6 75.Fe4 Fe8 76.Re5 Fb5 77.Fd5 Fe8 78.Re6 
  +-  (3.47)   Profondeur: 17/19   00:00:00  39kN, tb=1
69.Ff7 Fd7 70.Fd5 Fb5 71.Re6 Fe8 72.Fe4 Rd4 73.Ff3 Rc5 74.Fd5 Fg6 75.Fe4 Fe8 76.Re5 Fb5 77.Fd5 Fe8 78.Re6 
  +-  (3.47)   Profondeur: 17/19   00:00:00  39kN, tb=1
69.Ff7 Fd7 70.Fd5 Fb5 71.Re6 Fe8 72.Fe4 Rd4 73.Ff3 Rc5 74.Fd5 Fg6 75.Fe4 Fe8 76.Re5 Fb5 77.Fd5 Fe8 78.Re6 
  +-  (3.47)   Profondeur: 17/19   00:00:00  39kN, tb=1
69.Ff7 Fd7 70.Fd5 Fb5 71.Re6 Fe8 72.Fe4 Rd4 73.Ff3 Rc5 74.Fd5 Fg6 75.Fe4 Fe8 76.Re5 Fb5 77.Fd5 Fe8 78.Re6 
  +-  (3.47)   Profondeur: 17/19   00:00:00  39kN, tb=1
69.Ff7 Fd7 70.Fd5 Fb5 71.Re6 Fe8 72.Fe4 Rd4 73.Ff3 Rc5 74.Fg2 Fb5 75.Ff1 Fe8 76.Fd3 Rc6 77.Fe4+ Rc7 78.Fd5 Fg6 79.Fe4 
  +-  (3.47)   Profondeur: 18/20   00:00:00  49kN, tb=1
69.Ff7 Fd7 70.Fd5 Fb5 71.Re6 Fe8 72.Fe4 Rd4 73.Ff3 Rc5 74.Fg2 Fb5 75.Ff1 Fe8 76.Fd3 Rc6 77.Fe4+ Rc7 78.Fd5 Fg6 79.Fe4 
  +-  (3.47)   Profondeur: 19/20   00:00:00  53kN, tb=1
69.Ff7 Fd7 70.Fd5 Fb5 71.Re6 Fe8 72.Fe4 Rd4 73.Ff3 Rc5 74.Fg2 Fb5 75.Ff1 Fe8 76.Fd3 Rc6 77.Fe4+ Rc7 78.Fd5 Fg6 79.Fe4 
  +-  (3.47)   Profondeur: 20/20   00:00:00  60kN, tb=1
69.Ff7 Fd7 70.Fd5 Fb5 71.Re6 Fe8 72.Fe4 Rd4 73.Ff3 Rc5 74.Fg2 Fb5 75.Fe4 Rd4 76.Fd5 Rc5 77.Fe4 
  +-  (3.55)   Profondeur: 21/24   00:00:00  85kN, tb=1
69.Ff7 Fd7 70.Fd5 Fb5 71.Re6 Fe8 72.Fe4 Rd4 73.Ff3 Rc5 74.Fg2 Fb5 75.Fe4 Fe8 76.Ff3 Fa4 77.Fd5 Fe8 
  +-  (3.49)   Profondeur: 22/25   00:00:00  112kN, tb=1
69.Ff7 Fd7 70.Fd5 Fb5 71.Re6 Fe8 72.Fe4 Rd4 73.Ff3 Rc5 74.Fg2 Fb5 75.Fe4 Fe8 76.Ff3 Fa4 77.Re5 Fe8 78.Re6 
  +-  (3.47)   Profondeur: 23/25   00:00:00  129kN, tb=1
69.Ff7 Fd7 70.Fd5 Fb5 71.Re6 Fe8 72.Fe4 Rd4 73.Ff3 Rc5 74.Fg2 Fb5 75.Fe4 Fe8 76.Ff3 Fa4 77.Fh5 Fb5 78.Ff7 Fc4+ 79.Rf6 Fb5 80.Fg6 Rd6 81.Rf7 Fc6 82.Ff5 Fd5+ 83.Rf8 Fc6 84.Rf7 
  +-  (3.49)   Profondeur: 24/29   00:00:00  162kN, tb=1
69.Ff7 Fd7 70.Fd5 Fb5 71.Re6 Fe8 72.Fe4 Rd4 73.Ff3 Rc5 74.Fg2 Fb5 75.Fe4 Fe8 76.Ff3 Fa4 77.Fa8 Fb5 78.Fe4 
  +-  (3.41)   Profondeur: 25/29   00:00:00  314kN, tb=1
69.Ff7 Fd7 70.Fd5 Fb5 71.Re6 Fe8 72.Fe4 Rd4 73.Ff3 Rc5 74.Fg2 Fb5 75.Fe4 Fa4 76.Fg6 Fc6 77.Re5 Fd7 78.Fe4 Fb5 79.Fh1 Fe8 80.Re6 Fb5 81.Fe4 
  +-  (3.41)   Profondeur: 26/29   00:00:00  367kN, tb=1
69.Re6 Fc6 70.Fg6 Fa4 71.Rf7 Rd6 72.Fe4 Re5 73.Fc2 Fd7 74.Fd1 Fc6 75.Fb3 Rd6 76.Rf8 Fd7 77.Ff7 Re5 78.Fh5 Fa4 79.Fd1 Fb5 80.Fb3 Fd7 81.Fd1 Rd6 82.Fh5 Re5 83.Ff7 Fa4 84.Fh5 Fc6 
  +-  (3.47)   Profondeur: 27/29   00:00:00  480kN, tb=1
69.Re6 Fc6 70.Fg6 Fa4 71.Rf7 Rd6 72.Fe4 Re5 73.Fc2 Fd7 74.Fd1 Fc6 75.Fb3 Rd6 76.Rf8 Fd7 77.Ff7 Re5 78.Fh5 Rd6 79.Fe8 Ff5 80.Fa4 Fg6 81.Fb3 Rc6 82.Ff7 
  +-  (3.49)   Profondeur: 28/29   00:00:00  541kN, tb=1
69.Re6 Fc6 70.Fg6 Fa4 71.Rf7 Rd6 72.Fe4 Re5 73.Fc2 Fd7 74.Fd1 Fc6 75.Fb3 Rd6 76.Rf8 Fd7 77.Ff7 Re5 78.Fh5 Fb5 79.Fe8 Fe2 80.Ff7 Fb5 81.Fb3 Fc6 82.Ff7 Fa4 83.Fb3 Fd7 84.Ff7 Fb5 
  +-  (3.55)   Profondeur: 29/35   00:00:00  642kN, tb=1 
Why not 0.00 ? and why only 1 tb access ??

P.S. : I check when position have more than 5 pieces, it works correctly !
Edmund
Posts: 670
Joined: Mon Dec 03, 2007 3:01 pm
Location: Barcelona, Spain

Re: New 6-piece tablebases

Post by Edmund »

Vinvin wrote:Why not 0.00 ? and why only 1 tb access ??
If I am not mistaken it is called swindle mode. If the root position is available for probing it will elimate bad moves and then play the best move from an alpha beta search, putting pressure on the opponent and luring him to make a mistake.
User avatar
velmarin
Posts: 1600
Joined: Mon Feb 21, 2011 9:48 am

Re: New 6-piece tablebases

Post by velmarin »

Edmund wrote:
Vinvin wrote:Why not 0.00 ? and why only 1 tb access ??
If I am not mistaken it is called swindle mode. If the root position is available for probing it will elimate bad moves and then play the best move from an alpha beta search, putting pressure on the opponent and luring him to make a mistake.

I have also downloaded most of the torrent.
I like the project, but it raises questions, at least in the subject Stockfish, this is not correct.

Analysis of Comstock with Robbobases (600 megs + -), a perfect eval.
There are things to improve.

Image