Funny fortress position

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

Moderators: hgm, Rebel, chrisw

zamar
Posts: 613
Joined: Sun Jan 18, 2009 7:03 am

Funny fortress position

Post by zamar »

In Stockfish's self-play there occured a funny fortress position:

[D]4kb2/2p3r1/1pP1P2Q/pP2K3/P7/8/8/8 w - - 0 93

Stockfish shows this as +4 for white. How about your favourite engine?
Joona Kiiski
Top
User avatar
Michael Diosi
Posts: 672
Joined: Mon Jun 22, 2009 1:37 pm

Re: Funny fortress position

Post by Michael Diosi »

Hi,

Code: Select all

FEN: 4kb2/2p3r1/1pP1P2Q/pP2K3/P7/8/8/8 w - - 0 93 

SOS 5.1 for Arena:
  1/6	00:00	         261	0	 0,00	1.Ke5d4
  2/6	00:00	         263	0	 0,00	1.Ke5d4
  3/6	00:00	         302	0	 0,00	1.Ke5d4
  4/6	00:00	         349	0	 0,00	1.Ke5d4
  5/8	00:00	         430	0	 0,00	1.Ke5d4
  6/8	00:00	         566	0	 0,00	1.Ke5d4
  7/10	00:00	         824	0	 0,00	1.Ke5d4
  8/12	00:00	       1.218	0	 0,00	1.Ke5d4
  9/12	00:00	       1.848	0	 0,00	1.Ke5d4
 10/14	00:00	       2.737	0	 0,00	1.Ke5d4
 11/14	00:00	       4.155	0	 0,00	1.Ke5d4
 12/16	00:00	       6.374	0	 0,00	1.Ke5d4
 13/18	00:00	       8.951	0	 0,00	1.Ke5d4
 14/19	00:00	      12.789	0	 0,00	1.Ke5d4
 15/19	00:00	      15.351	0	 0,00	1.Ke5d4
 16/19	00:00	      18.358	0	 0,00	1.Ke5d4
 17/20	00:00	      24.107	0	 0,00	1.Ke5d4
 18/21	00:00	      30.473	0	 0,00	1.Ke5d4
 19/22	00:00	      38.049	0	 0,00	1.Ke5d4
 20/25	00:00	      48.119	0	 0,00	1.Ke5d4
 21/25	00:00	      66.896	0	 0,00	1.Ke5d4
 22/26	00:00	      89.106	0	 0,00	1.Ke5d4
 23/27	00:00	     106.307	0	 0,00	1.Ke5d4
 24/29	00:00	     125.364	0	 0,00	1.Ke5d4
 25/29	00:00	     145.447	0	 0,00	1.Ke5d4
 26/31	00:00	     215.778	686.486	 0,00	1.Ke5d4
 27/32	00:00	     260.489	727.626	 0,00	1.Ke5d4
 28/34	00:00	     349.975	727.626	 0,00	1.Ke5d4
 29/35	00:00	     462.275	727.626	 0,00	1.Ke5d4
 30/36	00:00	     535.874	727.626	 0,00	1.Ke5d4
 31/38	00:00	     660.128	853.776	 0,00	1.Ke5d4
 32/39	00:01	   1.000.371	922.096	 0,00	1.Ke5d4
 33/39	00:01	   1.180.612	951.719	 0,00	1.Ke5d4
 34/40	00:01	   1.410.094	995.865	 0,00	1.Ke5d4
 35/42	00:01	   1.620.128	995.865	 0,00	1.Ke5d4
 36/45	00:03	   3.093.339	1.027.409	 0,00	1.Ke5d4
 37/45	00:03	   3.422.497	1.031.360	 0,00	1.Ke5d4
 38/49	00:03	   3.823.888	1.031.360	 0,00	1.Ke5d4
 39/49	00:05	   5.767.867	1.042.189	 0,00	1.Ke5d4
 40/49	00:07	   7.466.502	1.029.771	 0,00	1.Ke5d4
 41/49	00:09	   9.487.588	1.051.700	 0,00	1.Ke5d4
 42/52	00:12	  13.347.112	1.058.706	 0,00	1.Ke5d4
 43/58	00:16	  17.565.283	1.050.856	 0,00	1.Ke5d4
 44/58	00:21	  21.853.164	1.004.230	 0,00	1.Ke5d4
Michael
http://www.playwitharena.com
Top
zamar
Posts: 613
Joined: Sun Jan 18, 2009 7:03 am

Re: Funny fortress position

Post by zamar »

Michael Diosi wrote:Hi,

Code: Select all

FEN: 4kb2/2p3r1/1pP1P2Q/pP2K3/P7/8/8/8 w - - 0 93 

SOS 5.1 for Arena:
  1/6	00:00	         261	0	 0,00	1.Ke5d4

[/quote]

Vow! How on earth it's able to spot this???
Joona Kiiski
Top
User avatar
rvida
Posts: 481
Joined: Thu Apr 16, 2009 12:00 pm
Location: Slovakia, EU

Re: Funny fortress position

Post by rvida »

zamar wrote:
Michael Diosi wrote:Hi,

Code: Select all

FEN: 4kb2/2p3r1/1pP1P2Q/pP2K3/P7/8/8/8 w - - 0 93 

SOS 5.1 for Arena:
  1/6	00:00	         261	0	 0,00	1.Ke5d4

[/quote]

Vow! How on earth it's able to spot this???[/quote]

SOS has special code to detect blocked positions.
Top
metax
Posts: 344
Joined: Wed Sep 23, 2009 5:56 pm
Location: Germany

Re: Funny fortress position

Post by metax »

Code: Select all

4kb2/2p3r1/1pP1P2Q/pP2K3/P7/8/8/8 w - -

Engine: ChessMind (0 MB)
von metax

6/24   0:01   +7.58    1.Dh5+ Kd8 2.Kf6 Te7 3.Dd5+ Kc8 
                       4.Df5 Kb8 5.Ke5 118 

7/21   0:01   +7.58    1.Dh5+ Kd8 2.Kf6 Te7 3.Dd5+ Ke8 
                       4.Dh5+ Kd8 166 

8/24   0:01   +7.58    1.Dh5+ Kd8 2.Kf6 Te7 3.Dd5+ Ke8 
                       4.Dh5+ Kd8 (342.292) 181 

9/25   0:02   +7.58    1.Dh5+ Kd8 2.Kf6 Te7 3.Dd5+ Ke8 
                       4.Dh5+ Kd8 (519.371) 190 

10/28  0:04   +7.58    1.Dh5+ Kd8 2.Kf6 Te7 3.Dd5+ Ke8 
                       4.Dh5+ Kd8 (851.105) 196 

11/29  0:06   +7.58    1.Dh5+ Kd8 2.Kf6 Te7 3.Dd5+ Ke8 
                       4.Dh5+ Kd8 (1.296.347) 203 

12/32  0:10   +7.58    1.Dh5+ Kd8 2.Kf6 Te7 3.Dd5+ Ke8 
                       4.Dh5+ Kd8 (2.150.164) 200 

13/36  0:22   +7.58    1.Dh5+ Kd8 2.Kf6 Te7 3.Dd5+ Ke8 
                       4.Dh5+ Kd8 (4.523.416) 201 

14/41  0:39   +7.58    1.Dh5+ Kd8 2.Kf6 Te7 3.Dd5+ Ke8 
                       4.Dh5+ Kd8 (8.266.660) 208 

14/41  0:51   +7.68    1.Kd5 Kd8 2.Dh4+ Te7 3.Dh8 Ke8 
                       4.Dh5+ Kd8 5.Dh8 Ke8 6.Dh5+ Kd8 
                       7.Dh8 Ke8 8.Dh5+ Kd8 9.Dh8 Ke8 
                       10.Dh5+ (10.898.019) 209 

15/37  1:03   +7.68    1.Kd5 Kd8 2.Dh4+ Te7 3.Dh8 Ke8 
                       4.Dh5+ Kd8 5.Dh8 (13.363.736) 209
:oops: :oops: :oops:

I don't know how this evaluation is reached, but it is clearly too extreme. :D
Top
metax
Posts: 344
Joined: Wed Sep 23, 2009 5:56 pm
Location: Germany

Re: Funny fortress position

Post by metax »

Thanks for this position, I just discovered some severe bugs in my evaluation function. :)

For example, each black pawn was only worth 0.90 pawns (and 0.75 pawns in the endgame) actually because I did the following in my pawn structure evaluation:

Code: Select all

  if (pawnCount[i].Black == 1)
           {
                  middlegame_value += 10;
                  endgame_value += 25;
           }
These should actually be the penalties for isolated pawn, but I forgot to test for the isolation condition. :D
A nice +80 for white with all pawns on the board, and even more in the endgame. :D
Top
metax
Posts: 344
Joined: Wed Sep 23, 2009 5:56 pm
Location: Germany

Re: Funny fortress position

Post by metax »

And yet another bug: The pawn C6 is, according to my evaluation, a "protected passed pawn" which gets a huge bonus. :D

I still have to find out why, but anyway thanks for this position, Joona. ;)

edit: Fixed. And yet another +10 Elo. :)
Top
metax
Posts: 344
Joined: Wed Sep 23, 2009 5:56 pm
Location: Germany

Re: Funny fortress position

Post by metax »

OK, I got it 'right' now. Still a very high evaluation, but better than +7.7... Maybe the king position including king safety is a bit overvalued with +1 for white.

Code: Select all

4kb2/2p3r1/1pP1P2Q/pP2K3/P7/8/8/8 w - -

Engine: ChessMind (0 MB)
von metax

8/23   0:01   +4.91    1.Dh5+ Kd8 2.Kf6 Te7 3.Dd5+ Ke8 
                       4.Dh5+ Kd8 (275.862) 213 

9/24   0:01   +4.91    1.Dh5+ Kd8 2.Kf6 Te7 3.Dd5+ Ke8 
                       4.Dh5+ Kd8 (446.471) 225 

10/27  0:03   +4.91    1.Dh5+ Kd8 2.Kf6 Te7 3.Dd5+ Ke8 
                       4.Dh5+ Kd8 (836.924) 239 

11/32  0:05   +4.91    1.Dh5+ Kd8 2.Kf6 Te7 3.Dd5+ Ke8 
                       4.Dh5+ Kd8 (1.375.159) 242 

12/31  0:08   +4.91    1.Dh5+ Kd8 2.Kf6 Te7 3.Dd5+ Ke8 
                       4.Dh5+ Kd8 (2.103.235) 244 

13/33  0:13   +4.91    1.Dh5+ Kd8 2.Kf6 Te7 3.Dd5+ Ke8 
                       4.Dh5+ Kd8 (3.298.288) 246 

14/38  0:22   +4.91    1.Dh5+ Kd8 2.Kf6 Te7 3.Dd5+ Ke8 
                       4.Dh5+ Kd8 (5.411.749) 245 

15/40  0:38   +4.91    1.Dh5+ Kd8 2.Kf6 Te7 3.Dd5+ Ke8 
                       4.Dh5+ Kd8 (9.551.982) 245 

16/36  0:50   +4.91    1.Dh5+ Kd8 2.Kf6 Te7 3.Dd5+ Ke8 
                       4.Dh5+ Kd8 (12.364.223) 245
PS: Sorry for my quadruple post. :)
Top
Edmund
Posts: 670
Joined: Mon Dec 03, 2007 3:01 pm
Location: Barcelona, Spain

Re: Funny fortress position

Post by Edmund »

rvida wrote:SOS has special code to detect blocked positions.
Has anyone got any ideas how this could be done? I know about corresponding squares, but this doesn't apply here with the pieces adding complexity.

regards,
Edmund
Top
User avatar
rvida
Posts: 481
Joined: Thu Apr 16, 2009 12:00 pm
Location: Slovakia, EU

Re: Funny fortress position

Post by rvida »

Edmund wrote:
rvida wrote:SOS has special code to detect blocked positions.
Has anyone got any ideas how this could be done? I know about corresponding squares, but this doesn't apply here with the pieces adding complexity.

regards,
Edmund
An excelent paper on the subjetct by Omid Tabibi, Ariel Felner and Nathan Netanyahu:
http://www.ise.bgu.ac.il/faculty/felner ... ckagej.pdf
Top

Return to “Computer Chess Club: General Topics”