Sharp & Knife edge positions

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

Martin Brown
Posts: 46
Joined: Sun Oct 18, 2009 12:07 pm

Sharp & Knife edge positions

Post by Martin Brown »

Greetings. This is my first post here having come from Usenet so please forgive any minor transgressions. I am interesting in analysing puzzle positions and have seen some very odd behaviour in engines. I am hoping that someone here may be able to elucidate why these things happen. The positions are tested under multipv mode with all lines displayed and infinite analysis from the start position.

The position that re-awoke my curiousity was posted last week by Tantale as a thread in r.g.c.analysis titled "A very sharp position". It is a difficult win for white but no engines can see it without human help - and the two most famous engines Fritz and Rybka cannot even do that reliably. Shredder10 and Crafty 20.14 both manage it with a bit of human help.

8/7p/1p4pk/4K3/5PP1/2p3r1/4p3/2R5 w - - 0 1

The rough plan is reasonably obvious to a human. It hinges on double Zugswang. I won't spoil it for anyone that wants to try and solve it by posting the solution here, but the full solution is online at:

http://www.jmrw.com/Chess/France/base.htm
Last study "Joachim Iglesias (2)

To help them along I played what I though would be the right lines and Shredder suddenly got it. HIARCS fails to find the solution but gets to the required depth in a couple of hours. Fritz8 could not get it even with 6 plys played out to help it. Rybka232 also gets stuck. The basic search looked OK at first with successive deepenings taking 2-3x longer out to 16 ply at 9hrs and it then spent over 50 hours considering the ramifications of Ra5 (already solved as #9) without making any further progress! I was expecting it to solve it completely in about that time.

I think that the problem is related to that seen in other knife edge positions where a wrong move by either side creates a quick loss. The resulting mate #N seems to mess up the iterative deepening search when the ply depth goes >2N and I don't understand why.

I have constructed a much simpler example where engines running in multipv mode under the Chessbase GUI quickly grind to a standstill when analysing the already *solved* lines. eg.

k6K/P1p3rP/qQP3Pb/1PP3p1/8/8/8/8 w - - 0 1

1. (#1) 1. Qb8# instant
2. (#3) 1. Qxa6 Rxh7+ 2. Kxh7 Bf8 3. Qb7# takes >7 mins at ply 20
3. (7.51) 1. bx6 cxb6 2. cxb6 Rxg6 3.c7 Bg7+ etc, takes 3s
4. (-15.69) 1. Qb7+ Qxb7 2. cxb7 Kxa7 3. b6+ Kb8 takes 3s
5. (-#3) 1. Qxc7 Rxc7 2. g7 Qa1 3. > 10 mins
6. (-#1) 1. Qa5 Qc8# > 7 mins

So at ply 20 it spends 24 minutes on solved lines and 6s on useful work. Only the mate #1 for the player to move behaves as expected. It is unclear why. Thanks for any enlightenment or suggestions for engines that are optimised for endgame and puzzle composing/solution.

Best regards,
Martin Brown
Martin Brown
Top
BBauer
Posts: 658
Joined: Wed Mar 08, 2006 8:58 pm

Re: Sharp & Knife edge positions

Post by BBauer »

Perhaps you should use a better program. :D
Here is what I got from stockfish1.5.1
FEN: 8/7p/1p4pk/4K3/5PP1/2p3r1/4p3/2R5 w - - 0 1

Stockfish_151_x64_ja:

Code: Select all

FEN: 8/7p/1p4pk/4K3/5PP1/2p3r1/4p3/2R5 w - - 0 1 

Stockfish_151_x64_ja:
   6	00:00	      12.333	59.579	-5,93	Ke5d4 c3c2 Rc1h1+ Kh6g7 g4g5 b6b5 Rh1e1 h7h5 g5xh6/ep+ Kg7xh6 Re1h1+ Kh6g7
   7	00:00	      18.501	86.051	-5,97	Ke5d4 c3c2 Rc1h1+ Kh6g7 g4g5 Rg3a3 Rh1e1 h7h5 g5xh6/ep+ Kg7xh6 Re1h1+ Kh6g7
   8-	00:00	      35.725	158.777	-6,34	Ke5d4 c3c2 Rc1a1 Rg3f3
   9-	00:00	      75.118	310.404	-7,83	Ke5d4 c3c2 Rc1a1 Rg3f3
  10	00:00	     361.003	1.022.671	-10,74	Ke5d4 Rg3f3 Rc1h1+ Kh6g7 Rh1c1 Rf3f1 Kd4d3 Rf1xc1 Kd3xe2 h7h5 g4xh5 g6xh5
  10	00:00	     401.238	1.069.968	-8,12	Ke5f6 Rg3h3 Rc1e1 c3c2 Re1c1 b6b5 Rc1e1 b5b4 Kf6e5 Rh3e3+ Ke5d4 c2c1Q Re1xc1 e2e1Q Rc1xe1 Re3xe1
  10	00:00	     432.746	1.103.943	-8,00	Ke5e4 c3c2 f4f5 g6xf5+ g4xf5 Kh6g5 Rc1e1 Rg3c3 Re1c1 Kg5f6 Ke4d4 e2e1Q Rc1xe1 c2c1Q Re1xc1 Rc3xc1
  11	00:00	     576.843	1.194.291	-8,24	Ke5e4 c3c2 Rc1e1 Rg3c3 Re1h1+ Kh6g7 Rh1c1 b6b5 Ke4d4 b5b4 f4f5 e2e1R Rc1xe1 c2c1Q Re1xc1 Rc3xc1
  12	00:00	     695.865	1.293.429	-8,80	Ke5e4 c3c2 Rc1e1 Rg3c3 Re1h1+ Kh6g7 Rh1c1 b6b5 Ke4d4 b5b4 f4f5 Kg7f6 f5xg6 Kf6xg6 g4g5 Kg6xg5
  12	00:00	     768.434	1.331.774	-7,59	Ke5f6 Rg3h3 Rc1e1 c3c2 Re1a1 Rh3h4 Ra1e1 b6b5 Kf6e5 Rh4xg4 Re1h1+ Kh6g7 Ke5e4 Rg4h4 Rh1e1
  13	00:00	     840.037	1.381.639	-7,07	Ke5f6 Rg3h3 Rc1e1 c3c2 Re1a1 Rh3h4 Ra1e1 b6b5 Kf6e5 Rh4xg4 Ke5e4 Rg4g3 Re1h1+ Kh6g7 Rh1e1
  14+	00:00	   1.619.968	1.838.783	-5,37	Ke5f6 Rg3h3 Kf6f7 Rh3h2 Kf7g8 b6b5 Rc1e1 b5b4 Re1a1 e2e1R Ra1xe1 Rh2h4 Kg8f7
  15	00:01	   1.969.688	1.936.763	-4,52	Ke5f6 Rg3h3 Kf6f7 Rh3h4 Rc1e1 c3c2 Kf7g8 g6g5 Re1xe2 g5xf4 Re2xc2 Kh6g5 Rc2f2 Kg5xg4 Kg8f7 f4f3 Kf7e6
  16	00:01	   2.255.762	1.977.004	-4,84	Ke5f6 Rg3h3 Kf6f7 Rh3h4 Rc1e1 c3c2 Kf7g8 g6g5 Re1xe2 g5xf4 Re2xc2 Kh6g5 Kg8f7 f4f3 Kf7e6 Rh4xg4 Ke6e5 h7h5
  17	00:01	   2.705.080	2.080.830	-4,80	Ke5f6 Rg3h3 Kf6f7 Rh3h4 Rc1e1 c3c2 Kf7g8 g6g5 Re1xe2 g5xf4 Re2xc2 Kh6g5 Kg8f7 f4f3 Kf7e6 Kg5xg4 Ke6e5 Kg4g3 Rc2c1
  18	00:01	   3.475.417	2.234.994	-4,68	Ke5f6 Rg3h3 Kf6f7 Rh3h4 Rc1e1 c3c2 Re1c1 g6g5 Rc1xc2 g5xf4 Rc2xe2 Kh6g5 Re2f2 Kg5xg4 Rf2g2+ Kg4f5 Rg2f2 Rh4h3 Kf7e7 h7h5 Rf2c2
  19+	00:01	   4.283.337	2.348.320	-4,40	Ke5f6 Rg3h3 Kf6f7 Rh3h4 Rc1e1 c3c2 Kf7f6 b6b5 Re1a1 b5b4 Kf6f7 c2c1R Ra1xc1 Rh4h3 Kf7g8 b4b3 Rc1c7
  20+	00:02	   6.552.527	2.514.400	-3,15	Ke5f6 Rg3h3 Rc1e1 c3c2 Kf6f7 Rh3h4 Kf7f6 b6b5 Re1a1 b5b4 Kf6f7 c2c1R Ra1xc1 b4b3 Kf7g8 Rh4h3 Rc1c7
  21	00:07	  20.554.772	2.802.668	-2,10	Ke5f6 Rg3h3 Rc1e1 Rh3h2 Kf6f7 g6g5 f4f5 Rh2h4 Re1xe2 Rh4xg4 Re2c2 Rg4c4 Rc2h2+ Rc4h4 Rh2c2 g5g4 f5f6 g4g3 Rc2xc3 Rh4g4 Kf7e8 g3g2 f6f7 Rg4e4+ Ke8d7 Kh6g7 Rc3g3+ Kg7xf7 Rg3xg2 h7h5 Kd7d6
  22	00:07	  22.065.763	2.809.493	-2,30	Ke5f6 Rg3h3 Rc1e1 Rh3h2 Kf6f7 g6g5 f4f5 Rh2h4 Re1xe2 Rh4xg4 Re2c2 Rg4c4 Rc2h2+ Rc4h4 Rh2c2 g5g4 f5f6 g4g3 Rc2xc3 Rh4g4 Kf7e8 g3g2 f6f7 Rg4e4+ Ke8d7 Kh6g7 Rc3g3+ Kg7xf7 Rg3xg2 h7h5 Kd7d6
  23+	00:09	  26.782.295	2.852.822	-1,69	Ke5f6 Rg3h3 Rc1a1 Rh3h4 Ra1b1 b6b5 Rb1a1 b5b4 Kf6f7 b4b3 Kf7g8 e2e1R Ra1xe1 Rh4h5 Re1e7
  24+	00:12	  34.919.164	2.887.790	+0,92	Ke5f6 Rg3h3 Rc1a1 Rh3h4 Ra1b1 b6b5 Rb1a1 b5b4 Kf6f7 b4b3 Kf7g8 e2e1R Ra1xe1 Rh4h5 Re1e7
  25	00:51	 148.540.830	2.861.782	+6,94	Ke5f6 Rg3h3 Rc1a1 Rh3h4 Ra1b1 c3c2 Rb1a1 Rh4h5 Kf6f7 Rh5f5+ g4xf5 Kh6h5 f5f6 Kh5g4 Kf7e6 Kg4f3 f6f7 Kf3f2 f7f8Q e2e1Q+ Ra1xe1 Kf2xe1 Qf8b4+ Ke1d1 Qb4d4+ Kd1e1 Qd4c3+ Ke1d1 Qc3d3+ Kd1c1 Ke6e5 Kc1b1 Qd3b3+ Kb1c1 Ke5e4 Kc1d2 Qb3d3+ Kd2c1
  26	00:57	 164.754.136	2.865.388	+7,43	Ke5f6 Rg3h3 Rc1a1 Rh3h4 Ra1b1 c3c2 Rb1a1 Rh4h5 Kf6f7 Rh5f5+ g4xf5 Kh6h5 f5f6 Kh5g4 Kf7e6 Kg4f3 f6f7 Kf3f2 f7f8Q e2e1Q+ Ra1xe1 Kf2xe1 Qf8b4+ Ke1d1 Qb4d4+ Kd1e1 Qd4c3+ Ke1d1 Qc3d3+ Kd1c1 Ke6e5 Kc1b2 Qd3d4+ Kb2b1 Qd4xb6+ Kb1a2 Qb6c5 Ka2b3 Ke5e4
  27+	01:44	 289.413.469	2.772.212	+11,79	Ke5f6
  28	02:46	 451.064.593	2.708.478	+15,23	Ke5f6 Rg3h3 Rc1a1 Rh3h4 Ra1b1 c3c2 Rb1a1 Rh4h5 Kf6f7 Rh5f5+ g4xf5 Kh6h5 f5f6 Kh5g4 f4f5 g6g5 Kf7g7 Kg4g3 f6f7 Kg3f2 f7f8Q e2e1Q Ra1xe1 Kf2xe1 Qf8b4+ Ke1d1 Qb4d4+ Kd1e2 Qd4c3 Ke2d1 Qc3d3+ Kd1c1 f5f6 Kc1b2 Qd3d4+ Kb2a2 Qd4c3 Ka2b1 Qc3b3+ Kb1c1 Qb3xb6 Kc1d2
  29	03:31	 575.116.421	2.718.673	+15,83	Ke5f6 Rg3h3 Rc1a1 Rh3h4 Ra1b1 c3c2 Rb1a1 Rh4h5 Kf6f7 Rh5f5+ g4xf5 Kh6h5 f5f6 Kh5g4 f4f5 g6g5 Kf7g7 Kg4g3 f6f7 Kg3f2 f7f8Q e2e1Q Ra1xe1 Kf2xe1 Qf8b4+ Ke1d1 Qb4d4+ Kd1e2 Qd4c3 Ke2d1 Qc3d3+ Kd1c1 f5f6 Kc1b2 Qd3d4+ Kb2b1 Qd4xb6+ Kb1a1 Qb6a5+ Ka1b2 Qa5e5+ Kb2b3 Qe5xg5 h7h5
  30	03:58	 653.794.674	2.738.957	+15,79	Ke5f6 Rg3h3 Rc1a1 Rh3h4 Ra1b1 c3c2 Rb1a1 Rh4h5 Kf6f7 Rh5f5+ g4xf5 Kh6h5 f5f6 Kh5g4 f4f5 g6g5 Kf7g7 Kg4g3 f6f7 Kg3f2 f7f8Q e2e1Q Ra1xe1 Kf2xe1 Qf8b4+ Ke1d1 Qb4d4+ Kd1e2 Qd4c3 Ke2d1 Qc3d3+ Kd1c1 f5f6 Kc1b2 Qd3d4+ Kb2b3 Qd4xb6+ Kb3c3 Qb6e3+ Kc3b2 Qe3e5+ Kb2b3 Qe5xg5 h7h5
  31+	05:39	 932.104.077	2.747.008	+16,48	Ke5f6
kind regards
Bernhard
Top
Martin Brown
Posts: 46
Joined: Sun Oct 18, 2009 12:07 pm

Re: Sharp & Knife edge positions

Post by Martin Brown »

StockFish 1.51 is decidedly impressive on this position and very fast too. How does it do that? Where do I get a legal copy?

But I was rather hoping one of the wizards here could explain to me what is happening to the multipv search mode in the other position that causes things to grind to a standstill when considering lines where there is a clear mate in N with a search depth >2N.
Martin Brown
Top
brianr
Posts: 536
Joined: Thu Mar 09, 2006 3:01 pm

Re: Sharp & Knife edge positions

Post by brianr »

I would not read too much into a single position. Small diferences in evaluation can easily trigger much faster or slower solutions.
Not sure what you mean by help, but running Tinker on this and "helping" for the first two moves, which it finds anyway, the score jumps to a clearly winning line with the 3rd move after about 2min. Of course, in any real game, Crafty and Rybka and StockFish all crush Tinker 99 times out of 100. So I think it is best to look at many positions with a few engines, or many engines for a few positions.

Code: Select all

ply    score   time     nodes  pv
 11->   -484   0.53   1332725  1.  e5f6  g3h3  2.  c1a1  c3c2  3.  f6f7  h3h4
                               4.  a1e1  b6b5  5.  e1a1  b5b4  6.  a1g1

Search alpha1 restarting after x -514
 12     -572   1.18   3074497  1.  e5f6  g3h3  2.  c1a1  c3c2  3.  a1e1  b6b5
                               4.  e1a1  b5b4  5.  f6f7  h3h4
 12->   -572   1.62   4255763  1.  e5f6  g3h3  2.  c1a1  c3c2  3.  a1e1  b6b5
                               4.  e1a1  b5b4  5.  f6f7  h3h4

Search beta1 restarting after x -542
 13     -565   1.95   5071875  1.  e5f6  g3h3  2.  f6f7  c3c2  3.  f7g8  h3h4
                               4.  g8h8  b6b5  5.  c1a1  h4h3  6.  g4g5  h6h5
                               7.  h8h7  b5b4
 13->   -565   2.63   6851406  1.  e5f6  g3h3  2.  f6f7  c3c2  3.  f7g8  h3h4
                               4.  g8h8  b6b5  5.  c1a1  h4h3  6.  g4g5  h6h5
                               7.  h8h7  b5b4

Search beta1 restarting after x -535
 14     -453   3.37   8727101  1.  e5f6  g3h3  2.  f6f7  c3c2  3.  f7g8  h3h4
                               4.  c1a1  g6g5  5.  f4f5  e2e1=q  6.  a1e1
                              h4g4  7.  e1h1  g4h4  8.  h1c1
 14->   -453   4.77  12484718  1.  e5f6  g3h3  2.  f6f7  c3c2  3.  f7g8  h3h4
                               4.  c1a1  g6g5  5.  f4f5  e2e1=q  6.  a1e1
                              h4g4  7.  e1h1  g4h4  8.  h1c1

Search beta1 restarting after x -423
 15     -380   7.16  18641792  1.  e5f6  g3h3  2.  f6f7  c3c2  3.  f7g8  h3h4
                               4.  c1a1  g6g5  5.  f4f5  e2e1=q  6.  a1e1
                              h4g4  7.  e1c1  g4c4  8.  c1h1  c4h4
 15->   -380   9.34  24707989  1.  e5f6  g3h3  2.  f6f7  c3c2  3.  f7g8  h3h4
                               4.  c1a1  g6g5  5.  f4f5  e2e1=q  6.  a1e1
                              h4g4  7.  e1c1  g4c4  8.  c1h1  c4h4

Search beta1 restarting after x -350
 16     -275  12.55  33763242  1.  e5f6  g3h3  2.  f6f7  h3h4  3.  f7g8  g6g5
                               4.  f4f5  h4g4  5.  f5f6  g4f4  6.  f6f7  h6g6
                               7.  f7f8=Q  f4f8  8.  g8f8  h7h5  9.  c1e1
 16->   -275  18.64  50062844  1.  e5f6  g3h3  2.  f6f7  h3h4  3.  f7g8  g6g5
                               4.  f4f5  h4g4  5.  f5f6  g4f4  6.  f6f7  h6g6
                               7.  f7f8=Q  f4f8  8.  g8f8  h7h5  9.  c1e1
 17    17/18 35.02            c1h1
Search alpha1 restarting after x -305

Search alpha2 restarting after x -245
 17     -237  35.80  97838224  1.  e5f6  g3h3  2.  f6f7  h3h4  3.  f7g8  g6g5
                               4.  f4f5  h4g4  5.  f5f6  g4f4  6.  f6f7  h6g6
                               7.  f7f8=Q  f4f8  8.  g8f8  h7h5  9.  c1e1
                              c3c2
 17->   -237  35.83  97838256  1.  e5f6  g3h3  2.  f6f7  h3h4  3.  f7g8  g6g5
                               4.  f4f5  h4g4  5.  f5f6  g4f4  6.  f6f7  h6g6
                               7.  f7f8=Q  f4f8  8.  g8f8  h7h5  9.  c1e1
                              c3c2
 18     -247  37.97 103856193  1.  e5f6  g3h3  2.  f6f7  h3h4  3.  f7g8  g6g5
                               4.  f4f5  h4g4  5.  f5f6  g4f4  6.  f6f7  g5g4
                               7.  f7f8=Q  f4f8  8.  g8f8  h6g5  9.  c1e1
                              c3c2  10.  e1a1  g4g3
 18->   -247  51.74 142075204  1.  e5f6  g3h3  2.  f6f7  h3h4  3.  f7g8  g6g5
                               4.  f4f5  h4g4  5.  f5f6  g4f4  6.  f6f7  g5g4
                               7.  f7f8=Q  f4f8  8.  g8f8  h6g5  9.  c1e1
                              c3c2  10.  e1a1  g4g3
 19     -239  58.54 161375324  1.  e5f6  g3h3  2.  f6f7  h3h4  3.  f7g8  g6g5
                               4.  f4f5  h4g4  5.  f5f6  g4f4  6.  f6f7  g5g4
                               7.  f7f8=Q  f4f8  8.  g8f8  h6g5  9.  c1e1
                              c3c2  10.  e1a1  g4g3  11.  f8e7
 19->   -239   1:29 248062394  1.  e5f6  g3h3  2.  f6f7  h3h4  3.  f7g8  g6g5
                               4.  f4f5  h4g4  5.  f5f6  g4f4  6.  f6f7  g5g4
                               7.  f7f8=Q  f4f8  8.  g8f8  h6g5  9.  c1e1
                              c3c2  10.  e1a1  g4g3  11.  f8e7

go
ply    score   time     nodes  pv
 15      244   0.49   1535349  1.  ...   g3h3  2.  f6f7
 15->    244   0.50   1535432  1.  ...   g3h3  2.  f6f7
 16      244   0.94   2912860  1.  ...   g3h3  2.  f6f7
 16->    244   0.95   2912943  1.  ...   g3h3  2.  f6f7
 17      244   1.78   5549905  1.  ...   g3h3  2.  f6f7
 17->    244   1.80   5549988  1.  ...   g3h3  2.  f6f7
 18      244   3.61  11189634  1.  ...   g3h3  2.  f6f7
 18->    244   4.31  13082991  1.  ...   g3h3  2.  f6f7

go
ply    score   time     nodes  pv
 15     -244   0.42   1169603  2.  f6f7  g6g5  3.  f4f5  h3h4  4.  f7g8  h4g4
 15->   -244   0.56   1521610  2.  f6f7  g6g5  3.  f4f5  h3h4  4.  f7g8  h4g4
 16     -244   1.17   3320053  2.  f6f7  g6g5  3.  f4f5  h3h4  4.  f7g8  h4g4
 16->   -244   1.59   4462076  2.  f6f7  g6g5  3.  f4f5  h3h4  4.  f7g8  h4g4
 17     -244   3.09   8707172  2.  f6f7  g6g5  3.  f4f5  h3h4  4.  f7g8  h4g4
 17->   -244   4.59  12951248  2.  f6f7  g6g5  3.  f4f5  h3h4  4.  f7g8  h4g4
 18     -244   9.22  25272878  2.  f6f7  h3h4  3.  f7g8
 18->   -244  14.07  39151854  2.  f6f7  h3h4  3.  f7g8
 19     1/21 25.01            f6f7
Search beta1 restarting after x -214
 19     -225  53.10 145592693  2.  f6f7  g6g5  3.  f4f5  h3h4  4.  c1c3  h4g4
                               5.  c3e3  g4c4  6.  e3e6  h6h5  7.  e6e2  g5g4
                               8.  f5f6  h5h6  9.  f7e6  c4c6  10.  e6e7  h6g6
                               11.  e2f2  g4g3  12.  f6f7  g3f2  13.  f7f8=Q
 19    10/21 1:15            c1a1
Root researching
 19       ??   1:17 208563585  2.  c1a1
 19      536   2:16 386070009  2.  c1a1  h3h4  3.  a1b1  c3c2  4.  b1a1  h4h5
 19->    536   2:25 415909367  2.  c1a1  h3h4  3.  a1b1  c3c2  4.  b1a1  h4h5
Top
Edmund
Posts: 670
Joined: Mon Dec 03, 2007 3:01 pm
Location: Barcelona, Spain

Re: Sharp & Knife edge positions

Post by Edmund »

Martin Brown wrote:StockFish 1.51 is decidedly impressive on this position and very fast too. How does it do that? Where do I get a legal copy?

But I was rather hoping one of the wizards here could explain to me what is happening to the multipv search mode in the other position that causes things to grind to a standstill when considering lines where there is a clear mate in N with a search depth >2N.
One factor definitly is Mate Distance Pruning.
Also discussed here: http://www.talkchess.com/forum/viewtopic.php?t=26995

Some engines have it, others don't. It doesn't add to playing strength as it only comes into affect if the position is already decided.
Top
Vinvin
Posts: 5228
Joined: Thu Mar 09, 2006 9:40 am
Full name: Vincent Lejeune

Re: Sharp & Knife edge positions

Post by Vinvin »

You can get it here : http://homepages.tesco.net/henry.ablett/jims.html
Martin Brown wrote:StockFish 1.51 is decidedly impressive on this position and very fast too. How does it do that? Where do I get a legal copy?
Top
Martin Brown
Posts: 46
Joined: Sun Oct 18, 2009 12:07 pm

Re: Sharp & Knife edge positions

Post by Martin Brown »

Thanks. There are a lot of other engines new to me on that site.
Are they all freely available to download and play with?
Martin Brown
Top
Martin Brown
Posts: 46
Joined: Sun Oct 18, 2009 12:07 pm

Re: Sharp & Knife edge positions

Post by Martin Brown »

Edmund wrote:
Martin Brown wrote:StockFish 1.51 is decidedly impressive on this position and very fast too. How does it do that? Where do I get a legal copy?

But I was rather hoping one of the wizards here could explain to me what is happening to the multipv search mode in the other position that causes things to grind to a standstill when considering lines where there is a clear mate in N with a search depth >2N.
One factor definitly is Mate Distance Pruning.
Also discussed here: http://www.talkchess.com/forum/viewtopic.php?t=26995


Some engines have it, others don't. It doesn't add to playing strength as it only comes into affect if the position is already decided.
Thanks for this. Yes it explains the slowdown I have seen. Pity a few more engines do not include mate distance pruning when they are running in analysis mode. It is really annoying watching it waste an hour on a solved line and then spend a minute on the only unresolved one.

Can anyone recommend a strong engine that includes this feature?
Martin Brown
Top
Edmund
Posts: 670
Joined: Mon Dec 03, 2007 3:01 pm
Location: Barcelona, Spain

Re: Sharp & Knife edge positions

Post by Edmund »

Martin Brown wrote:Thanks for this. Yes it explains the slowdown I have seen. Pity a few more engines do not include mate distance pruning when they are running in analysis mode. It is really annoying watching it waste an hour on a solved line and then spend a minute on the only unresolved one.

Can anyone recommend a strong engine that includes this feature?
I think Glaurung has it, so probably Stockfish too.
Top

Return to “Computer Chess Club: Programming and Technical Discussions”