7-men Syzygy attempt

Discussion of chess software programming and technical issues.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
syzygy
Posts: 4458
Joined: Tue Feb 28, 2012 10:56 pm

Re: 7-men Syzygy attempt

Post by syzygy » Sun Jun 10, 2018 1:23 pm

noobpwnftw wrote:
Sun Jun 10, 2018 12:49 pm
syzygy wrote:
Sun Jun 10, 2018 12:15 pm
noobpwnftw wrote:
Sun Jun 10, 2018 11:55 am


337 tables to go, progress of 5v2 pawnful (141/210) & 4v3 pawnful (57/325).
I just noticed KRBPPvKQ is missing. Was there any problem with its generation?
KRBBPvKQ is missing somehow, retrying, maybe OOM or a simple mis-hit of Ctrl+C. :D
Make sure you type KRBPP and not KRBBP ;-)

syzygy
Posts: 4458
Joined: Tue Feb 28, 2012 10:56 pm

Re: 7-men Syzygy attempt

Post by syzygy » Sun Jun 10, 2018 1:26 pm

syzygy wrote:
Sun Jun 10, 2018 1:23 pm
noobpwnftw wrote:
Sun Jun 10, 2018 12:49 pm
syzygy wrote:
Sun Jun 10, 2018 12:15 pm

I just noticed KRBPPvKQ is missing. Was there any problem with its generation?
KRBBPvKQ is missing somehow, retrying, maybe OOM or a simple mis-hit of Ctrl+C. :D
Make sure you type KRBPP and not KRBBP ;-)
Oops, that was not a typo by yours. Indeed the subtable KRBBPvKQ is missing, which explains why KRBPPvKQ failed.

noobpwnftw
Posts: 360
Joined: Sun Nov 08, 2015 10:10 pm

Re: 7-men Syzygy attempt

Post by noobpwnftw » Mon Jun 11, 2018 11:48 am

Nope, KRBBPvKQ needs a 16-bit table...

Code: Select all

Longest win for white: 100 ply; 5q2/P5R1/8/k7/B7/6B1/8/K7 w - -
Longest cursed win for white: 656 ply; 6B1/3B4/5R2/6q1/P7/1k6/8/3K4 b - -
Longest cursed win for black: 138 ply; 7B/8/1K6/7B/P7/4q1k1/8/7R w - -
Longest win for black: 60 ply; 2qk4/8/8/P4K2/8/8/RB6/4B3 w - -


Resetting black cursed losses.
time taken =   0:13.477
time taken =   0:13.610
time taken =   0:17.004
Very large DTZ not yet supported.tb_size = 146762325360

syzygy
Posts: 4458
Joined: Tue Feb 28, 2012 10:56 pm

Re: 7-men Syzygy attempt

Post by syzygy » Mon Jun 11, 2018 1:46 pm

noobpwnftw wrote:
Mon Jun 11, 2018 11:48 am
Nope, KRBBPvKQ need a 16-bit table...

Code: Select all

Longest win for white: 100 ply; 5q2/P5R1/8/k7/B7/6B1/8/K7 w - -
Longest cursed win for white: 656 ply; 6B1/3B4/5R2/6q1/P7/1k6/8/3K4 b - -
Longest cursed win for black: 138 ply; 7B/8/1K6/7B/P7/4q1k1/8/7R w - -
Longest win for black: 60 ply; 2qk4/8/8/P4K2/8/8/RB6/4B3 w - -


Resetting black cursed losses.
time taken =   0:13.477
time taken =   0:13.610
time taken =   0:17.004
Very large DTZ not yet supported.tb_size = 146762325360
Wow, that's unexpected.
If you want to continue the generation of the other tables while I work on 16-bit compression for pawnful tables, you could generate just KRBBPvKQ.rtbw for now (sufficient to generate KRBPPvKQ) and leave .rtbz for later.

Sesse
Posts: 180
Joined: Mon Apr 30, 2018 9:51 pm
Contact:

Re: 7-men Syzygy attempt

Post by Sesse » Mon Jun 11, 2018 1:54 pm

656 ply before being able to move the pawn safely… wow.

syzygy
Posts: 4458
Joined: Tue Feb 28, 2012 10:56 pm

Re: 7-men Syzygy attempt

Post by syzygy » Mon Jun 11, 2018 10:47 pm

Sesse wrote:
Mon Jun 11, 2018 1:54 pm
656 ply before being able to move the pawn safely… wow.
It seemed reasonably safe to assume that 580 ply would be enough for everybody (at least those with pawns).

The 656 ply is 656 ply to move the pawn to a win or 556 ply to move the pawn to another cursed win. (Or to a capture or mate, but that seems less likely.) We'll have to wait until we have the DTZ table to find out.

According to the Lomonosov tables, the longest mate in KRBBPvKQ is 799 moves, so knowing that, the 656 ply is maybe less surprising. I don't know what that longest mate looks like (but obviously it will have several pawn moves).

syzygy
Posts: 4458
Joined: Tue Feb 28, 2012 10:56 pm

Re: 7-men Syzygy attempt

Post by syzygy » Mon Jun 11, 2018 11:31 pm

syzygy wrote:
Mon Jun 11, 2018 10:47 pm
According to the Lomonosov tables, the longest mate in KRBBPvKQ is 799 moves,
Those must be 799 plies, not moves.

The position is apparently this:
[D]7R/4q3/8/BK6/8/k6P/8/B7 w - - 0 1

Arpad Rusz
Posts: 270
Joined: Sat Apr 17, 2010 12:34 pm
Location: Budapest

Re: 7-men Syzygy attempt

Post by Arpad Rusz » Sat Jun 16, 2018 10:47 am

I have started to develop a software for mining the new 7-man Syzygy tablebases.
http://ruszchessstudies.blogspot.com/20 ... r-see.html

For now, it is just a command line tool (actually it is a Stockfish engine with some additional UCI commands) but I plan to add a specialized GUI to the project as well.

A funny thing: Stockfish will have to learn how to generate retrograde moves as well for some of the planned features.

User avatar
Nordlandia
Posts: 2471
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: 7-men Syzygy attempt

Post by Nordlandia » Wed Jun 20, 2018 11:09 am

Is KNNPvKQP the most relevant set needed when analysing the "Behting Study" :?:

8/8/7p/3KNN1k/2p4p/8/3P2p1/8 w - -

Sesse
Posts: 180
Joined: Mon Apr 30, 2018 9:51 pm
Contact:

Re: 7-men Syzygy attempt

Post by Sesse » Thu Jun 21, 2018 5:34 pm

OK, I've extended the 7-man tablebase volume from 7 to 8 TB; after that, I can't reserve any more disk space for this. New disks (I have to throw out some old ones and start a new RAID) are about €2k, so we'll see when and if it comes to that. I guess the end estimate hasn't changed all that much.

Post Reply