mgv77 the four hard ones

[Marc MP]

Is there an engine that definitely chooses the proposed best move and holds it for some time? These seem to be especially difficult.

Code: Select all

8/3bkpp1/pp2p3/1P1pP2r/P2K3P/3B4/2P2P2/R7 w - -                  bm a5; hmvc 0; fmvc 1; id "MGV22-E: MOROZEVICH - Vallejo Pons, 1997";
8/5p2/r1k3p1/7p/2PK1P1P/1R4P1/8/8 w - -                          bm c5; hmvc 0; fmvc 1; id "GER - GER";
r2qkb1r/3bpppp/p4nn1/3P4/1p1N4/1QN3B1/PP3PPP/R3KB1R w KQkq -     bm Nc6; hmvc 0; fmvc 1; id "MGV35-P: SOLOVJOV - Kharitonov 2005";
r3r1k1/2qb1pp1/pp2n1np/2p1P3/2R1N3/1B3N1P/PP1Q1PP1/4R1K1 w - -   bm Nf6+; hmvc 0; fmvc 1; id "MGV70-K: NUNN - Smeets Amsterdam 2006";

[d]8/3bkpp1/pp2p3/1P1pP2r/P2K3P/3B4/2P2P2/R7 w - -

Nice surprise! Pro Deo 1.6 - default on my PIV 3.2GHZ

Code: Select all

00:00:00.5	 -0,10	1	bxa6 Rxh4+ Ke3 Rxa4 
00:00:00.5	  0,09	1	Rh1 
00:00:00.5	  0,08	2	Rh1 axb5 axb5 
00:00:00.6	  0,11	3	Rh1 axb5 axb5 Be8 
00:00:00.6 -0,50	4      Rh1 axb5 Be2 Rf5 
00:00:00.6	 -0,50	4	bxa6 
00:00:00.6	  0,28	4	bxa6 Rxh4+ Ke3 Bc6 
00:00:00.6	  0,23	5	bxa6 Rxh4+ Kc3 d4+ Kc4 Bc6 
00:00:00.6	  0,34	6      bxa6 Rxh4+ Kc3 d4+ Kc4 Bc6 Rg1 
00:00:00.6	  0,76	7	bxa6 Rxh4+ Kc3 d4+ Kc4 Bc6 Rb1 Bd5+ Kb5 Rf4 Kxb6 Rxf2 

00:00:00.7  0,28	8      bxa6 Rxh4+ Ke3 Rxa4 Rxa4 Bxa4 a7 Bc6 Bb5 Ba8 f4 

00:00:00.9  0,31	9	bxa6 Rxh4+ Kc3 Rxa4 Rxa4 Bxa4 Kb4 Bc6 Bb5 Ba8 Kc3 g5 

00:00:01.1	  0,29	10	bxa6 Rxh4+ Kc3 Rxa4 Rxa4 Bxa4 Kb4 Bc6 Bb5 Ba8 Kc3 Kd8 Kd4 Kc7 

00:00:01.8	  0,29	11    bxa6 Rxh4+ Kc3 Rxa4 Rxa4 Bxa4 Kb4 Bc6 Bb5 Ba8 Kc3 Kd8 Kd4 Kc7 

00:00:02.6	  0,23	12	bxa6 Rxh4+ Kc3 Rxa4 Rxa4 Bxa4 Kb4 Bc6 Bb5 Ba8 Kc3 Kd8 Kd4 Kc7 Be8 f6 

00:00:04.9	  0,23	13	bxa6 Rxh4+ Kc3 Rxa4 Rxa4 Bxa4 Kb4 Bc6 Bb5 Ba8 Kc3 Kd8 Kd4 Kc7 Be8 f6 

00:00:11.4	  0,30	14	bxa6 Rxh4+ Kc3 Rxa4 Rxa4 Bxa4 Kb4 Bc6 Bb5 Ba8 Kc3 Kd8 Kd4 Kc7 Be8 f6 Bf7 Kd7 Bg6 

00:00:27.0	  0,14	15	bxa6 Rxh4+ Kc3 Rxa4 Rxa4 f6 exf6+ Kxf6 Kb4 Bxa4 Kxa4 g5 a7 g4 

00:01:13.8	  0,14	16	bxa6 Rxh4+ Kc3 Rxa4 Rxa4 Bxa4 Kb4 Bc6 Bb5 Ba8 Kc3 Kd8 Kd4 Kc7 Be8 f5 Bf7 Kd7 f4 Bc6 a7 Ba8 Bh5 Bb7 

00:03:23.2	  0,22	17	bxa6 Rxh4+ Kc3 Rxa4 Rxa4 Bxa4 Kb4 Bc6 Bb5 Ba8 Kc3 Kd8 f4 Kc8 Kd4 Kc7 Be8 f6 Bb5 Bc6 a7 Ba8 Be2 Bb7 

00:08:13.7    0,20	18	bxa6 Rxh4+ Kc3 Rxa4 Rxa4 Bxa4 Kb4 Bc6 Bb5 Bd7 Bxd7 Kxd7 a7 f5 

00:15:54.2	   0,20	18	a5 

00:26:31.8   0,63	18	a5 Rxh4+ Kc3 bxa5 bxa6 a4 a7 Rh8 

On the same hardware, FireBird 1.2 with the 5 mens TribleBases after 7 hours:

Code: Select all

1.bxa6 Rxh4+ 2.Kc3 Bc6 3.a7 Kd7 4.a5 bxa5 5.Rxa5 Rh8 6.Kd4 Kc7 7.Kc5 Ba8 8.Rb5 Bb7 9.Rb4 g5 10.Rg4 Ra8 11.Rxg5 Rxa7 12.Rg7 Ra5+ 13.Kb4 Kb6 14.Rxf7 Bc8 15.Rf6 Ra2 16.f3 Kc7 
  ±  (1.28)   Depth: 33   07:36:32  15428mN, tb=38773706

I think this one might be broken actually:

1. a5 Rxh4+ 2. Kc3 Bxb5 3. Bxb5 axb5? 4. a6! wins easily.

[d]8/4kpp1/Pp2p3/1p1pP3/7r/2K5/2P2P2/R7 b - - 0 4

but 3... bxa5!...

[d]8/4kpp1/p3p3/pB1pP3/7r/2K5/2P2P2/R7 w - - 0 4

...followed at some point by Rf4 to snatch the f2 pawn (or... Re4 if white doesn't put a bishop in d3) and f6 to exchange the e5 pawn might well end up in a draw even if white is a bishop up.

It might have been influenced by the actual game play:

[Event "Pamplona International"]
[Site "Pamplona ESP"]
[Date "1999.01.03"]
[EventDate "?"]
[Round "07"]
[Result "1-0"]
[White "Alexander Morozevich"]
[Black "Francisco Vallejo-Pons"]
[ECO "C12"]
[WhiteElo "2625"]
[BlackElo "2450"]
[PlyCount "109"]

1. e4 e6 2. d4 d5 3. Nc3 Nf6 4. Bg5 Bb4 5. e5 h6 6. Bd2 Bxc3
7. bxc3 Ne4 8. Qg4 Kf8 9. Nf3 c5 10. Bd3 Nxd2 11. Kxd2 Nc6
12. Qf4 Qe7 13. dxc5 Qxc5 14. Nd4 Bd7 15. Rhb1 Nd8 16. a4 Rc8
17. Rb3 a6 18. h4 Rc7 19. g4 Nc6 20. Nxc6 Bxc6 21. Qb4 Qxb4
22. cxb4 Bd7 23. Rc3 Rxc3 24. Kxc3 Ke7 25. Kd4 h5 26. gxh5
Rxh5 27. b5 b6 28. a5 Rxh4+ 29. Kc3 Bxb5 30. Bxb5 axb5 31. a6
Rh8 32. Kb4 Kd7 33. Kxb5 Kc7 34. Ra3 f6 35. Rg3 Rh7 36. Rc3+
Kb8 37. Kxb6 Rh8 38. a7+ Ka8 39. exf6 gxf6 40. Rc7 Rd8 41. Rc6
Re8 42. f4 d4 43. Rd6 e5 44. f5 Rf8 45. Ka6 Rc8 46. Kb6 Rf8
47. Re6 Rd8 48. Kc7 Rf8 49. Kd7 Rf7+ 50. Ke8 Rc7 51. Rxf6 Rxc2
52. Rd6 e4 53. f6 e3 54. Rxd4 Rf2 55. f7 1-0

[Terry McCracken]

Is there an engine that definitely chooses the proposed best move and holds it for some time? These seem to be especially difficult.

Code: Select all

8/3bkpp1/pp2p3/1P1pP2r/P2K3P/3B4/2P2P2/R7 w - -                  bm a5; hmvc 0; fmvc 1; id "MGV22-E: MOROZEVICH - Vallejo Pons, 1997";
8/5p2/r1k3p1/7p/2PK1P1P/1R4P1/8/8 w - -                          bm c5; hmvc 0; fmvc 1; id "GER - GER";
r2qkb1r/3bpppp/p4nn1/3P4/1p1N4/1QN3B1/PP3PPP/R3KB1R w KQkq -     bm Nc6; hmvc 0; fmvc 1; id "MGV35-P: SOLOVJOV - Kharitonov 2005";
r3r1k1/2qb1pp1/pp2n1np/2p1P3/2R1N3/1B3N1P/PP1Q1PP1/4R1K1 w - -   bm Nf6+; hmvc 0; fmvc 1; id "MGV70-K: NUNN - Smeets Amsterdam 2006";

Especially Difficult for Engines??

I find this amazing as I found the *Key* Move 3 out of the 4 diagrams instantly. I made a mental slip for Nc6 but they appear quite easy.

Something the Human Mind excells at is patterns and intuition. Computers are not up to speed with humans recognizing patterns and intuition is still reserved for the human mind.

[Dann Corbit]

I am not sure that either move makes any progress. I suspect that the position is really drawn no matter what you do.

Two sample analysis records:

[d]8/3bkpp1/pp2p3/1P1pP2r/P2K3P/3B4/2P2P2/R7 w - - acd 34; acn 109963199249; acs 1065613; bm a5; ce 63; pv a5 Rxh4+ Kc3 Bxb5 Bxb5 bxa5 Bxa6 Rf4 Rxa5 f6 exf6+ gxf6 Kd3 Rxf2 Bc8 Rf3+ Kd2 Rf2+ Kd1 Rf1+ Ke2 Rf4 Ra7+ Kd8 Bd7 Ke7 Ba4+ Kd6 Ra6+ Ke7 Bb3 f5 Rb6 Re4+ Kf3 Rg4 Ke3 Rg3+ Kf4 Rg4+ Kf3 Rh4 Ra6 Rg4 Kf2 Rh4 Kg3 Rg4+ Kf2;

[d]8/3bkpp1/pp2p3/1P1pP2r/P2K3P/3B4/2P2P2/R7 w - - acd 34; acn 109963199249; acs 1065613; bm bxa6; ce 143; pv bxa6 Rxh4+ Ke3 Bc6 a5 Rh3+ Kd4 Rh4+ Kc3 bxa5 Rxa5 Rh8 Kb4 Kd7 a7 Kc7 Kc5 d4 Ra6 Bh1 Bb5 Ra8 Kxd4;

[Terry McCracken]

I am not sure that either move makes any progress. I suspect that the position is really drawn no matter what you do.

Two sample analysis records:

[d]8/3bkpp1/pp2p3/1P1pP2r/P2K3P/3B4/2P2P2/R7 w - - acd 34; acn 109963199249; acs 1065613; bm a5; ce 63; pv a5 Rxh4+ Kc3 Bxb5 Bxb5 bxa5 Bxa6 Rf4 Rxa5 f6 exf6+ gxf6 Kd3 Rxf2 Bc8 Rf3+ Kd2 Rf2+ Kd1 Rf1+ Ke2 Rf4 Ra7+ Kd8 Bd7 Ke7 Ba4+ Kd6 Ra6+ Ke7 Bb3 f5 Rb6 Re4+ Kf3 Rg4 Ke3 Rg3+ Kf4 Rg4+ Kf3 Rh4 Ra6 Rg4 Kf2 Rh4 Kg3 Rg4+ Kf2;

[d]8/3bkpp1/pp2p3/1P1pP2r/P2K3P/3B4/2P2P2/R7 w - - acd 34; acn 109963199249; acs 1065613; bm bxa6; ce 143; pv bxa6 Rxh4+ Ke3 Bc6 a5 Rh3+ Kd4 Rh4+ Kc3 bxa5 Rxa5 Rh8 Kb4 Kd7 a7 Kc7 Kc5 d4 Ra6 Bh1 Bb5 Ra8 Kxd4;

I took another look at this position and it appears White can't force the win with a5 the natural tactical shot which would normally win but Black isn't forced to play what's in the game and White's pawns are very weak and the extra piece will then only be useful to draw.

28. ba! is White's best try and I think it may win as the Black Rook will be forced back to the back rank to defend.

Try that and play around with it a bit.

[Paloma]

Hi Marc,

i think 1.bxa6! is better as 1.a5 (which leads in a drawish position.)

Here my try: 1.bxa6! Rxh4 2.Kc3 Rxa4 3.Rxa4 Bxa4 and now 4.a7! Bc6 5.Kd4 +-

e.g. 5...Kd8 6.f4 Kc8 7.Bb5 Ba8 8.Be8 f5! 9.Bb5 Kb7! 10.Bd7 Kxa7 11.Bxe6 Bc6

12.Bxf5 Kb7 13.e6 Kc7 14.Ke5 and white wins. A long interaktive analyses by me with different engines.

Terry thinks also 1.bxa6 is better.
http://www.talkchess.com/forum/viewtopi ... 43&t=33172

[Marc MP]

Hi Marc,

i think 1.bxa6! is better as 1.a5 (which leads in a drawish position.)

Here my try: 1.bxa6! Rxh4 2.Kc3 Rxa4 3.Rxa4 Bxa4 and now 4.a7! Bc6 5.Kd4 +-

e.g. 5...Kd8 6.f4 Kc8 7.Bb5 Ba8 8.Be8 f5! 9.Bb5 Kb7! 10.Bd7 Kxa7 11.Bxe6 Bc6

12.Bxf5 Kb7 13.e6 Kc7 14.Ke5 and white wins. A long interaktive analyses by me with different engines.

Terry thinks also 1.bxa6 is better.
http://www.talkchess.com/forum/viewtopi ... 43&t=33172

Hi Herbert,

I also think the line you give: 1.bxa6! Rxh4 2.Kc3 Rxa4 3.Rxa4 Bxa4 and now 4.a7! Bc6 5.Kd4 +- will eventually win for white. But black seems to have a better defencive try: 1.bxa6! Rxh4 2.Kc3 Bc6!

[d]8/4kpp1/Ppb1p3/3pP3/P6r/2KB4/2P2P2/R7 w - - 2 3

The engines I tried up to now are looking at 3.a7 or 3.Rb1 with scores in between +1.00 and +1.60, but nothing decisive yet.

[jdart]

Arasan (latest build) got #4 (Nunn-Smeets) but it took a long time.

"MGV70-K: NUNN - Smeets Amsterdam 2006" bm Nf6+
result: Nf6+ score: +1.40 ++ solved in 1408.03 sec. (4078.48M nodes)
Nf6+ gxf6
result(2): Nd6 score: +1.05 ** not solved in 2500.09 secs. (
7175.93M nodes)
Nd6 Bc6 Rg4 Bxf3 gxf3 Qd7 f4 Rad8 f5 Nd4 Rxd4
result(3): Rc3 score: +0.10 ** not solved in 2500.03 secs. (
7430.52M nodes)
Rc3 Bc6 Nd6 Re7 Qc2

[Dann Corbit]

Arasan (latest build) got #4 (Nunn-Smeets) but it took a long time.

"MGV70-K: NUNN - Smeets Amsterdam 2006" bm Nf6+
result: Nf6+ score: +1.40 ++ solved in 1408.03 sec. (4078.48M nodes)
Nf6+ gxf6
result(2): Nd6 score: +1.05 ** not solved in 2500.09 secs. (
7175.93M nodes)
Nd6 Bc6 Rg4 Bxf3 gxf3 Qd7 f4 Rad8 f5 Nd4 Rxd4
result(3): Rc3 score: +0.10 ** not solved in 2500.03 secs. (
7430.52M nodes)
Rc3 Bc6 Nd6 Re7 Qc2

I am convinced that the last two are correct, but I doubt the correctness of the first two.

[Marc MP]

Arasan (latest build) got #4 (Nunn-Smeets) but it took a long time.

"MGV70-K: NUNN - Smeets Amsterdam 2006" bm Nf6+
result: Nf6+ score: +1.40 ++ solved in 1408.03 sec. (4078.48M nodes)
Nf6+ gxf6
result(2): Nd6 score: +1.05 ** not solved in 2500.09 secs. (
7175.93M nodes)
Nd6 Bc6 Rg4 Bxf3 gxf3 Qd7 f4 Rad8 f5 Nd4 Rxd4
result(3): Rc3 score: +0.10 ** not solved in 2500.03 secs. (
7430.52M nodes)
Rc3 Bc6 Nd6 Re7 Qc2

I am convinced that the last two are correct, but I doubt the correctness of the first two.

[d]8/5p2/r1k3p1/7p/2PK1P1P/1R4P1/8/8 w - -

Here is FireBird 1.2 output with 3-4-5 Triblebases. I'm not sure the score is high enough so that 1.c5 must be considered like a win based on the engine evaluation, even if the PV seems winning. I've the feeling that the TribleBases are a bit buggy (I tried others positions as well). Maybe some one with Shredder loaded with bothe ShredderBases and 5 mens tablebases could find something clearer?

Code: Select all

<snip>
1.c5 Kc7 2.Kd5 Ra1 3.Re3 Kd8 4.c6 Rd1+ 5.Ke5 Ke7 6.Rc3 Re1+ 7.Kd5 Rd1+ 8.Kc5 Kd8 9.Rb3 Kc8 10.Ra3 Kc7 11.Ra7+ Kc8 12.Rxf7 Rc1+ 13.Kd6 Rd1+ 14.Ke7 Rg1 15.f5 gxf5 
  +-  (1.81)   Depth: 32   01:06:31  2784mN, tb=133238792

1.c5 Kc7 2.Kd5 Ra2 3.Re3 Kd8 4.c6 Ra1 5.Kd6 Rd1+ 
  +-  (1.89)   Depth: 32   01:48:47  4776mN, tb=202212846

1.c5 Kc7 2.Kd5 Ra2 3.Re3 Kd8 4.c6 Ra1 5.Kd6 Rd1+ 
  +-  (1.89)   Depth: 32   01:48:47  4776mN, tb=202212846

1.c5 Kc7 2.Kd5 Ra2 3.Re3 Kd8 4.c6 Rd2+ 5.Ke5 Ke7 6.Rc3 Re2+ 7.Kd5 Rd2+ 8.Kc5 Kd8 9.Ra3 Rc2+ 10.Kd6 Rd2+ 11.Ke5 Rc2 12.Kf6 Kc7 13.Kxf7 Rxc6 
  +-  (2.01)   Depth: 32   03:30:51  10505mN, tb=309270287

1.c5 Kc7 2.Kd5 Ra2 3.Re3 Kd8 4.c6 Rd2+ 5.Ke5 Ke7 6.Rc3 Re2+ 7.Kd5 Rd2+ 8.Kc5 Kd8 9.Ra3 Rc2+ 10.Kd6 Rd2+ 11.Ke5 Rc2 12.Kf6 Kc7 13.Kxf7 Rxc6 
  +-  (2.01)   Depth: 32   03:30:51  10505mN, tb=309270287

1.c5 Kc7 2.Kd5 Ra2 3.Re3 Kd8 4.c6 Rd2+ 5.Ke5 Ke7 6.Rc3 Re2+ 7.Kd5 Rd2+ 8.Kc5 Kd8 9.Ra3 Rc2+ 10.Kd6 Rd2+ 11.Ke5 Rc2 12.Kf6 Kc7 13.Kxf7 Rxc6 14.Re3 Kd7 15.Rd3+ 
  +-  (2.19)   Depth: 32   08:43:31  29694mN, tb=523512264

1.c5 Kc7 2.Kd5 Ra2 3.Re3 Kd8 4.c6 Rd2+ 5.Ke5 Ke7 6.Rc3 Re2+ 7.Kd5 Rd2+ 8.Kc5 Kd8 9.Ra3 Rc2+ 10.Kd6 Rd2+ 11.Ke5 Rc2 12.Kf6 Kc7 13.Kxf7 Rxc6 14.Re3 Kd7 15.Rd3+ 
  +-  (2.19)   Depth: 32   08:43:31  29694mN, tb=523512264