Page 1 of 2

Testposition Endgame

Posted: Sat Jul 17, 2010 10:33 pm
by Eduard Nemeth
This Testposition is created by me.

5b2/p2k1p2/P3pP1p/n2pP1p1/1p1P2P1/1P1KB2N/7P/8 w - -

1. Nxg5! ++-

How much time your program needs to find this move?

ED.

Re: Testposition Endgame

Posted: Sat Jul 17, 2010 11:51 pm
by John Doe
Eduard Nemeth wrote:
How much time your program needs to find this move?
.
Analysis by Stockfish 1.8 JA 64bit:

1.Kc2 Nc6 2.Kd3 Na5
=/+ (-0.44) Depth: 6 00:00:00 1kN
= (0.00) Depth: 45 00:01:09 452mN
1.Nxg5 hxg5
= (0.08 ++) Depth: 45 00:04:34 1864mN
+/= (0.32 ++) Depth: 45 00:08:15 3317mN

Re: Testposition Endgame

Posted: Sun Jul 18, 2010 6:19 am
by John Doe
Analysis by IvanHoe 9.53b x64:

...
1.Kc2 Nc6 2.Kd3 Na5
=/+ (-0.35 ++) Depth: 23 00:00:02 14045kN
= (0.00) Depth: 32 00:00:29 199mN
1.Nxg5 hxg5
= (0.08 ++) Depth: 32 00:07:10 3138mN
= (0.08 ++) Depth: 32 00:07:10 3138mN

Actually he found the move much faster but it took him a lot of time to output

Re: Testposition Endgame

Posted: Sun Jul 18, 2010 10:59 am
by Robert Houdart
Does Nxg5 actually win?
You cannot stop the analysis at scores of +0.32 or +0.08...

Re: Testposition Endgame

Posted: Sun Jul 18, 2010 8:05 pm
by Eduard Nemeth
Robert Houdart wrote:Does Nxg5 actually win?
You cannot stop the analysis at scores of +0.32 or +0.08...
Dear Robert!

Yes, 1. Nxg5 is the only move here that wins (1. Bxg5 is only draw).

Real position is this, with an comment by M. Gurevich from CSS Forum.
(He is not the GM M. Gurevich; he is an other player with ELO ~2200)

5b2/p2k1p2/P3pP1p/n2pP1p1/1p1P2P1/1P1KBN2/7P/8 w - - 0 1

[Event "Advanced Chess"]
[Site "CSS-Forum, WM-Test"]
[Date "2000.??.??"]
[Round "?"]
[White "Anand"]
[Black "Shirov"]
[Result "1-0"]
[Annotator "WM-Test Extra 55) Sf3xg5 (E)"]
[SetUp "1"]
[FEN "5b2/p2k1p2/P3pP1p/n2pP1p1/1p1P2P1/1P1KBN2/7P/8 w - - 0 1"]
[PlyCount "13"]
[EventDate "2000.??.??"]

1. Nxg5 $3 hxg5 2. Bxg5 Nxb3 3. h4 Na1 4. Bc1 Nb3 5. Be3 Na5 6. g5 Nc4 7. Bc1
$18 1-0

ED.

Re: Testposition Endgame

Posted: Sun Jul 18, 2010 8:25 pm
by Eduard Nemeth
How I see at moment, Houdini 1.03a have hard problems to find the winning move after the last move from the analysis.

Example:

5b2/p2k1p2/P3pP2/3pP1P1/1pnP3P/3K4/8/2B5 b - -

Engine: Houdini 1.03a w32 1-cpu (1024 MB)
von Robert Houdart

11/26 0:00 -0.27++ 7...b3 8.Kc3 La3 9.Lxa3 Sxa3 10.Kxb3 Sb5
11.h5 Sxd4+ 12.Kc3 Sf3 13.g6 d4+
14.Kc4 fxg6 15.h6 Sxe5+ 16.Kxd4 Sf3+
17.Ke4 (112.542) 1461

11/31 0:00 -0.48++ 7...b3 8.Kc3 La3 9.Lxa3 Sxa3 10.Kxb3 Sb5
11.h5 Sxd4+ 12.Kc3 Sf3 13.g6 d4+
14.Kb4 fxg6 15.hxg6 Ke8 16.g7 Kf7
17.Kc4 Kg8 18.Kd3 Sxe5+ 19.Kxd4 Sf7 (119.860) 1461

11/39 0:00 +0.29 7...b3 8.Kc3 b2 9.Lxb2 Lb4+ 10.Kxb4 Sxb2
11.h5 Sd3+ 12.Kc3 Sf4 13.h6 Ke8
14.h7 Sg6 15.Kc2 Kf8 16.Kd2 Sh8
17.Kd3 Sg6 18.Ke3 Sh8 19.Kf4 Sg6+
20.Kg4 (559.084) 1663

12/39 0:00 +0.29 7...b3 8.Kc3 b2 9.Lxb2 Lb4+ 10.Kxb4 Sxb2
11.h5 Sd3+ 12.Kc3 Sf4 13.h6 Ke8
14.h7 Sg6 15.Kc2 Kf8 16.Kd2 Sh8
17.Kd3 Sg6 18.Ke3 Sh8 19.Kf4 Sg6+
20.Kg4 (704.602) 1650

13/39 0:00 +0.29 7...b3 8.Kc3 b2 9.Lxb2 Lb4+ 10.Kxb4 Sxb2
11.h5 Sd3+ 12.Kc3 Sf4 13.h6 Ke8
14.h7 Sg6 15.Kc2 Kf8 16.Kd2 Sh8
17.Kd3 Sg6 18.Ke3 Sh8 19.Kf4 Sg6+
20.Kg4 (989.763) 1649

14/39 0:01 +0.29 7...b3 8.Kc3 b2 9.Lxb2 Lb4+ 10.Kxb4 Sxb2
11.h5 Sd3+ 12.Kb3 Sf4 13.h6 Sg6
14.h7 Kc6 15.Ka4 Kd7 16.Kb5 Sh8
17.Kc5 Sg6 18.Kb4 Kc7 19.Kb3 Kc6
20.Ka4 (3.252.410) 1672

15/39 0:02 +0.29 7...b3 8.Kc3 b2 9.Lxb2 Lb4+ 10.Kxb4 Sxb2
11.h5 Sd3+ 12.Kb3 Sf4 13.h6 Sg6
14.h7 Kc6 15.Ka4 Kd7 16.Kb5 Sh8
17.Kc5 Sg6 18.Kb4 Kd8 19.Kb5 Kd7
20.Kb4 Kd8 (3.434.562) 1677

16/39 0:02 +0.29 7...b3 8.Kc3 b2 9.Lxb2 Lb4+ 10.Kxb4 Sxb2
11.h5 Sd3+ 12.Kb3 Sf4 13.h6 Sg6
14.h7 Kc6 15.Ka4 Kd7 16.Kb5 Sh8
17.Kc5 Sg6 18.Kb4 Kd8 19.Kb5 Kd7
20.Kb4 Kd8 (4.333.756) 1679

17/39 0:03 +0.29 7...b3 8.Kc3 b2 9.Lxb2 Lb4+ 10.Kxb4 Sxb2
11.h5 Sd3+ 12.Kb3 Sf4 13.h6 Sg6
14.h7 Kc6 15.Ka4 Kd7 16.Kb5 Sh8
17.Kb4 Sg6 18.Kc5 Sh8 19.Kb5 Sg6
20.Kb4 Kc7 (5.879.536) 1676

18/39 0:05 +0.29 7...b3 8.Kc3 b2 9.Lxb2 Lb4+ 10.Kxb4 Sxb2
11.h5 Sd3+ 12.Kb5 Sf4 13.h6 Sg6
14.h7 Sh8 15.Kb4 Sg6 16.Kc5 Sh8
17.Kb5 Sg6 18.Kb4 Kc6 19.Ka5 Sh8
20.Ka4 Kc7 (8.703.289) 1676

19/60 2:01 +3.29 7...Lc5 8.h5 Lxd4 9.Kxd4 Sa5 10.h6 Sc6+
11.Kc5 Sxe5 12.Kxb4 d4 13.h7 Sg6
14.Kc4 d3 15.Kxd3 Kc6 16.Le3 Kb5
17.Lxa7 Kxa6 18.Lc5 Kb7 19.Ke4 Kc6
20.Lf8 Kd7 (202.030.418) 1667

What is wrong here by Houdini 1.03a? :shock: . Other engines find the winning very easy with an TOP-Score!

Example:

Crab 1.0 beta is here very fast with an TOP Score ~ +10!

5b2/p2k1p2/P3pP2/3pP1P1/1pnP3P/3K4/8/2B5 b - -

Engine: Crab 1.0 beta (1024 MB)
von Tord Romstad, Marco Costalba, Joona Kii

13.02 0:01 +0.20 7...b3 8.Kc3 Lb4+ 9.Kxb4 b2 10.Lxb2 Sxb2
11.h5 Sd3+ 12.Kb5 Sf4 13.h6 Sg6
14.h7 Sh8 15.Kc5 Sg6 16.Kb4 Ke8
17.Kb5 Kf8 18.Kc6 (437.404) 266

14.01 0:01 0.00 7...b3 8.Kc3 Lb4+ 9.Kxb4 b2 10.Lxb2 Sxb2
11.h5 Sd3+ 12.Kb5 Sf4 13.h6 Sg6
14.h7 Sh8 15.Kc5 Sg6 16.Kb4 Ke8
17.Kb5 Kd7 18.Kb4 (588.567) 342

15.01 0:01 0.00 7...b3 8.Kc3 Lb4+ 9.Kxb4 b2 10.Lxb2 Sxb2
11.h5 Sd3+ 12.Kb5 Sf4 13.h6 Sg6
14.h7 Sh8 15.Kc5 Sg6 16.Kb4 Ke8
17.Kb5 Kd7 18.Kb4 (739.961) 415

16.01 0:01 0.00 7...b3 8.Kc3 Lb4+ 9.Kxb4 b2 10.Lxb2 Sxb2
11.h5 Sd3+ 12.Kb5 Sf4 13.h6 Sg6
14.h7 Sh8 15.Kc5 Sg6 16.Kb4 Ke8
17.Kb5 Kd7 18.Kb4 (961.578) 508

17.01 0:02 0.00 7...b3 8.Kc3 Lb4+ 9.Kxb4 b2 10.Lxb2 Sxb2
11.h5 Sd3+ 12.Kb5 Sf4 13.h6 Sg6
14.h7 Sh8 15.Kc5 Sg6 16.Kb4 Ke8
17.Kb5 Kd7 18.Kb4 (1.207.132) 603

18.01 0:02 0.00 7...b3 8.Kc3 Lb4+ 9.Kxb4 b2 10.Lxb2 Sxb2
11.h5 Sd3+ 12.Kb5 Sf4 13.h6 Sg6
14.h7 Sh8 15.Kc5 Sg6 16.Kb4 Ke8
17.Kb5 Kd7 18.Kb4 (1.716.865) 779

19.01 0:02 0.00 7...b3 8.Kc3 b2 9.Lxb2 Lb4+ 10.Kxb4 Sxb2
11.h5 Sd3+ 12.Kb5 Sf4 13.h6 Sg6
14.h7 Sh8 15.Kc5 Sg6 16.Kb5 (2.420.198) 974

20.01 0:03 +0.08-- 7...b3 8.h5 La3 9.Lxa3 Sxa3 10.Kc3 Sc4
11.h6 b2 12.Kc2 b1S 13.Kxb1 (4.256.432) 1316

20.01 0:03 +0.16-- 7...b3 8.h5 Sb6 9.Kc3 Sa4+ 10.Kxb3 Sb6
11.h6 (4.764.427) 1386

20.01 0:03 +0.32-- 7...b3 8.h5 Lb4 9.h6 b2 10.Kc2 bxc1D+
11.Kxc1 Se3 12.h7 La3+ 13.Kd2 Sf1+
14.Kd3 (5.508.297) 1463

20.01 0:04 +0.64-- 7...b3 8.h5 b2 9.Lxb2 Sxb2+ 10.Kc2 Sc4
11.h6 Se3+ 12.Kc3 Sd1+ 13.Kd2 Sb2
14.h7 (6.534.614) 1566

20.01 0:04 +1.29-- 7...b3 8.h5 Lb4 9.h6 b2 10.Kc2 La3
11.Lxb2 Se3+ 12.Kd3 Lxb2 13.Kxe3 Lc1+
14.Kd3 Lxg5 15.h7 (7.917.387) 1672

20.01 0:05 +2.58-- 7...b3 8.h5 b2 9.Lxb2 Sxb2+ 10.Kc2 Sc4
11.h6 Se3+ 12.Kc3 Sd1+ 13.Kd2 Sb2
14.h7 Sc4+ 15.Kc1 La3+ 16.Kc2 Se3+
17.Kd3 (10.725.107) 1820

20.01 0:08 +5.17-- 7...b3 8.h5 b2 9.Lxb2 Lb4 10.Lc3 Lxc3
11.Kxc3 Ke8 12.h6 (17.473.665) 1975

20.01 0:16 +7.87 7...b3 8.h5 b2 9.Lxb2 Sxb2+ 10.Kc2 Sc4
11.h6 Lc5 12.h7 Lxd4 13.g6 fxg6
14.h8D Lxe5 15.Dg7+ Kd6 16.De7+ Kc6
17.Dxe6+ Ld6 18.f7 Kc5 19.Dxg6 Sa3+
20.Kd3 Sb5 (35.197.548) 2133

20.02 0:17 +5.91 7...Lc5 8.h5 Lxd4 9.Kxd4 Sa5 10.h6 Sc6+
11.Kc5 Sxe5 12.Kxb4 d4 13.Kc5 Sg6
14.Kxd4 Kc6 15.h7 Sh8 16.Ke4 Kb6
17.Le3+ Kxa6 18.Ke5 Kb7 19.Lc5 Kc6
20.Lxa7 (37.837.722) 2129

21.01 0:19 +4.56 7...Lc5 8.h5 Lxd4 9.Kxd4 Sa5 10.h6 Sc6+
11.Kc5 Sxe5 12.Kxb4 d4 13.Kc5 Sg6
14.Kxd4 Kc6 15.h7 Kb6 16.Le3 Kxa6
17.Kc5 Kb7 18.Kd6 a5 19.Lc5 a4
20.Lb4 Sh8 (41.217.573) 2137

22.01 0:20 +5.13 7...Lc5 8.h5 Lxd4 9.Kxd4 Sa5 10.h6 Sc6+
11.Kc5 Sxe5 12.Kxb4 d4 13.Kc5 Sg6
14.Kxd4 Kc6 15.h7 Kb6 16.Le3 Kb5
17.Ld2 Kxa6 18.Kc5 Kb7 19.Kd6 Kc8
20.La5 Kb7 (42.963.966) 2143

23.01 0:22 +6.38-- 7...Lc5 8.h5 b3 9.h6 b2 10.Lxb2 Sxb2+
11.Kc2 Lxd4 12.h7 Lxe5 13.h8D Ld4
14.g6 fxg6 15.Dg7+ Kc6 16.De7 Sa4
17.Dxe6+ Kc5 18.De7+ Kb6 19.f7 Lc5
20.Db7+ Ka5 (47.709.623) 2160

23.01 0:26 +7.63-- 7...Lc5 8.h5 Sa3 9.h6 Sc2 10.dxc5 Se1+
11.Ke2 (57.734.985) 2183

23.01 0:37 +9.37 7...Lc5 8.h5 b3 9.h6 b2 10.Lxb2 Sxb2+
11.Kc2 Lxd4 12.h7 Sc4 13.g6 fxg6
14.h8D Lxe5 15.Dh7+ Kc6 16.Db7+ Kc5
17.Dxa7+ Sb6 18.De7+ Ld6 19.Dxe6 g5
20.Kd3 g4 (81.620.032) 2194

What do you think about, Robert?

ED.

Re: Testposition Endgame

Posted: Sun Jul 18, 2010 10:23 pm
by Robert Houdart
Yes, an interesting case.
In the last position Houdini rapidly sees that 7...b3 loses (a "fail low" after approximately 10M nodes), but it takes another 200M positions to find a better move. It's so slow because Houdini goes very deep, "19/60" means that the maximum depth in some variations was 60 plies. Maybe some of the extensions should be cut down...

Note that after 1.Nxg5 hxg5 2.Bxg5 Nxb3 3.h4 Black can try 3...Bc5 4.dxc5 Nxc5+, it's maybe good enough for a draw, what do you think?

8/p2k1p2/P3pP2/2npP1B1/1p4PP/3K4/8/8 w - - 0 5

Robert

Re: Testposition Endgame

Posted: Sun Jul 18, 2010 11:43 pm
by hyatt
Eduard Nemeth wrote:This Testposition is created by me.

5b2/p2k1p2/P3pP1p/n2pP1p1/1p1P2P1/1P1KB2N/7P/8 w - -

1. Nxg5! ++-

How much time your program needs to find this move?

ED.
Crafty 36 seconds:

Code: Select all

               31    36.16     +1   1. Nxg5!                          
               31     1:07     +3   1. Nxg5!                          
               31     1:18   1.42   1. Nxg5 hxg5 2. Bxg5 Nxb3 3. h4 Na1
                                    4. Bd2 b3 5. Bc3 Nc2 6. g5 Ba3 7. h5
                                    Ne3 8. Kxe3 Bc1+ 9. Kd3 Bxg5 10. Bb4
                                    Kc6 11. Kc3 Kb5 12. Kxb3 Bh6 13. Kc3
                                    Bf4 14. Bf8 Kxa6 15. Bg7 Kb5 16. h6
                                    Kc6
So 36 seconds to find it,

Re: Testposition Endgame

Posted: Sun Jul 18, 2010 11:55 pm
by Eduard Nemeth
Hello Robert!

Thanks for your idea with 3...Bc5!?

Here is my analysis as PGN. In my opinion, white should wins anyway bit with a small way.
But 100% sure I am not.

Position after the last move of my analysis after 3...Bc5.

8/3k1p2/p3pP2/1K1p4/5B2/8/8/8 w - -

[Event "Advanced Chess [WM-Test Extra 55) Sf3xg5 (E)]"]
[Site "?"]
[Date "2000.??.??"]
[Round "?"]
[White "Anand"]
[Black "Shirov"]
[Result "1-0"]
[SetUp "1"]
[FEN "5b2/p2k1p2/P3pP1p/n2pP1p1/1p1P2P1/1P1KBN2/7P/8 w - -"]

1. Nxg5 hxg5 2. Bxg5 Nxb3 3. h4 Na1 (3... Bc5 4. dxc5 Nxc5+
5. Kc2 b3+ 6. Kb2 Nd3+ 7. Kxb3 Nxe5 8. h5 Kc6 9. Be3 Nxg4
10. h6 Nxh6 11. Bxh6 Kb5 12. Be3 Kxa6 13. Kb4 Kb7 14. Kc5
Kc7 15. Bf4+ Kd7 16. Kb5 a6+ (16... d4 17. Kc4) 17. Kc5)
4. Bc1 Nb3 5. Be3 Na5 6. g5 Nc4 7. Bc1 1-0

Crab 1.0 beta see here an big advantage for white.

32.01 0:08 +12.68-- 17.Kc5 a5 18.Kb5 a4 19.Kxa4 Kc6
20.Kb4 Kb6 21.Lg3 Kc6 22.Kc3 Kd7
23.Kd4 Kc6 24.Ke5 Kc5 25.Lf2+ Kc6
26.Kf4 Kd7 27.La7 Kc6 28.Kg5 Kd7
29.Lg1 Ke8 30.Ld4 (25.717.681) 3042

32.01 0:09 +77.08 17.Kc5 a5 18.Kb5 a4 19.Kxa4 Kc6
20.Kb4 Kb6 21.Lg3 Kc6 22.Kc3 Kd7
23.Kd4 Kc6 24.Ke5 Kc5 25.Lf2+ Kc6
26.Kf4 Kd7 27.La7 Kc6 28.Kg5 Kd7
29.Lg1 Ke8 30.Ld4 (27.425.530) 3047

33.01 0:09 +16.16 17.Kc5 a5 18.Kb5 a4 19.Kxa4 Kc6
20.Kb4 Kb6 21.Lg3 Kc6 22.Kc3 Kd7
23.Kd4 Kc6 24.Ke5 Kc5 25.Lf2+ Kc6
26.Kf4 Kd7 27.La7 Kc6 28.Kg5 Kd7
29.Kh6 e5 30.Kg7 (28.784.052) 3054

Bester Zug: Kb5-c5 Zeit: 0:11.781 min K/s: 3.105.123 Knoten: 36.581.464

What is your opinion?

ED.

Re: Testposition Endgame

Posted: Mon Jul 19, 2010 12:16 am
by Robert Houdart
3...Bc5 was the move suggested by Houdini.
In the line you give 10...Nxf6 appears to be an improvement for black.

1. Nxg5 hxg5 2. Bxg5 Nxb3 3. h4 Bc5 4. dxc5 Nxc5+ 5. Kc2 b3+ 6. Kb2 Nd3+ 7. Kxb3 Nxe5 8. h5 Kc6 9. Be3 Nxg4 10. h6 Nxf6

8/p4p2/P1k1pn1P/3p4/8/1K2B3/8/8 w - - 0 7

Houdini doesn't see a win for White in this position, nor does Stockfish. Obviously White has a positive score (+0.7 in Houdini, +2.3 in Stockfish), but it cannot be converted to a win.

If this is Anand v Shirov, Shirov may have missed a draw... :)