Rule of the square

220 views
Skip to first unread message

Mikael (mbootsector)

unread,
Sep 24, 2014, 3:14:42 PM9/24/14
to fishc...@googlegroups.com
I found this position 8/8/1k6/p7/5K2/5P1P/6P1/8 b - - 0 1 which stockfish evaluates as 1.89 for white.

Is there a reason why the rule of the square isn't in Stockfish? Has it been before?


Joseph Ellis

unread,
Sep 24, 2014, 3:31:04 PM9/24/14
to fishc...@googlegroups.com
I think this is a case of the problem being resolved easily enough in search.  

The search stays positive until depth 5, where it drops to -0.21, and at depth 6 it is -4.70 (0.01 seconds).

However, it should be trivial enough to add the rule to evaluate, if so desired... 

Ajedrecista

unread,
Sep 24, 2014, 3:32:25 PM9/24/14
to fishc...@googlegroups.com
El miércoles, 24 de septiembre de 2014 21:14:42 UTC+2, Mikael (mbootsector) escribió:
> I found this position 8/8/1k6/p7/5K2/5P1P/6P1/8 b - - 0 1 which stockfish evaluates as 1.89 for white.
>
>
> Is there a reason why the rule of the square isn't in Stockfish? Has it been before?

Hello Mikael:

I analysed this position with SF DD w32 (2 cores) under Arena 2.0.1 GUI (hash = 128 MB), and it fastly says 'win for the side to move' (evals over 85). You may analyse this position again, clearing the hash before or even re-starting the engine. If there is still wrong, there is a work to do: compare different compiles from Abrok site between SF DD and your version, until we spot where is the error.

Regards from Spain.

Ajedrecista.

Ajedrecista

unread,
Sep 24, 2014, 3:36:10 PM9/24/14
to fishc...@googlegroups.com
Hi again:

Sorry for my two consecutive posts but I think I have not been enough clear: when I said 'win for the side to move', I was referring to the fact that I analysed the position for both white and black to move, and in both cases the side to move wins according to SF DD, which is evident.

Regards from Spain.

Ajedrecista.

Mikael (mbootsector)

unread,
Sep 24, 2014, 4:01:53 PM9/24/14
to fishc...@googlegroups.com
Yes the solution is found quickly if you start from this position, but what if the position is found on previous searches after 30 ply? White might enter this line thinking he is winning when he is not. Given stockfishes heavy pruning you might need to search far deeper than 30 ply to find that this position is lost.

Uri Blass

unread,
Sep 24, 2014, 4:30:30 PM9/24/14
to fishc...@googlegroups.com
I think that there was some rule for unstoppable passed pawn but marco removed it because it was a long code that was not productive.

I think that having the square rule may be counter productive because there are cases when there is unstoppable passed pawn but the opponent has no passed pawn that can promote earlier so it may be better to have a huge bonus only when the unstoppable passed pawn is more advanced than all the opponent pawns so the opponent cannot defend by promoting faster with a pawn that is even not a passed pawn(for example white a2 black a3 b4 when black plays b3 and promote earlier.

Mikael (mbootsector)

unread,
Sep 25, 2014, 2:46:37 AM9/25/14
to fishc...@googlegroups.com
I see. Thanks.
Reply all
Reply to author
Forward
0 new messages