Incredible hard testposition Zukertort -Steinitz 1872

[rightrook]

>>>What's so special about 24...Bc6? I think 24...fxg5 wins too, and the difference between these moves is small. This is not a good test position<<<


Yes, I agree with Mike....

regards

Robert

[Ovyron]

I get different results every time I analyze a position. I guess that this is for two reasons:

It's because of Sampled Search, other GUIs don't support it and in Aquarium you can't switch it off.

[oreopoulos]

It is sampled search. In some positions it helps find smthg much faster, in others slower. Now go figure why. Noone (besides Vas) knows how this sampled search works

[Eelco de Groot]

I tried the testposition once more but Ancalagon keeps playing 24... fxg5, the eval does go down eventually but not enough to consider any other move and no forced draw is seen, at 19 ply finally 26.Qd3+ Kh6 27.h4 in the PV which is an alternative to directly playing 26. h4, Naum 3.1 played this 26.Qd3 also, although 26. h4 is better. 21 plies completed on the Athlon 2009 MHz:

[D]r2q4/pp1b2pp/5pk1/3Q2B1/8/8/PP4PP/4R2K b - -

Engine: Ancalagon 1.08 Testversion 002 (256 MB)
by Tord Romstad


9.00 0:00 +3.01 24...fxg5 25.Rd1 Rc8 26.Qd6+ Kh5
27.g4+ Kxg4 28.Qd4+ Kh5 29.Qxd7 Qxd7
30.Rxd7 Rc1+ 31.Kg2 Rc2+ 32.Kf3 g4+
33.Ke3 Rxb2 34.Rxg7 (253.569) 294

10.00 0:01 +3.01 24...fxg5 25.Rd1 Rc8 26.Qd6+ Kh5
27.g4+ Kxg4 28.Qd4+ Kh5 29.Qxd7 Qxd7
30.Rxd7 Rc1+ 31.Kg2 Rc2+ 32.Kf3 g4+
33.Ke3 Rxb2 34.Rxg7 (483.314) 355

11.01 0:02 +3.01 24...fxg5 25.Rd1 Rc8 26.Qd6+ Kh5
27.g4+ Kxg4 28.Qd4+ Kh5 29.Qxd7 Qxd7
30.Rxd7 Rc1+ 31.Kg2 Rc2+ 32.Kf3 g4+
33.Ke3 Rxb2 34.Rxg7 (1.125.901) 393

12.01 0:05 +3.01 24...fxg5 25.Rd1 Rc8 26.Qd6+ Kh5
27.g4+ Kxg4 28.Qd4+ Kh5 29.Qxd7 Qxd7
30.Rxd7 Rc1+ 31.Kg2 Rc2+ 32.Kf3 g4+
33.Ke3 Rxb2 34.Rxg7 (2.297.550) 410

13.01 0:12 +2.96 24...fxg5 25.Rd1 Rc8 26.Qd6+ Kh5
27.g4+ Kxg4 28.Qd4+ Kh5 29.Qxd7 Qxd7
30.Rxd7 Rc1+ 31.Kg2 Rc2+ 32.Kf3 Rxb2
33.Rxg7 Kh6 34.Re7 Rxh2 35.Rxb7 Rxa2 (5.200.907) 413

14.01 0:37 +2.98 24...fxg5 25.Rd1 Rc8 26.Qd6+ Kh5
27.g4+ Kxg4 28.Qd4+ Kh5 29.Qxd7 Qxd7
30.Rxd7 Rc1+ 31.Kg2 Rc2+ 32.Kf3 Rxb2
33.Rxg7 h6 34.h3 Rb4 (15.170.075) 403

15.01 1:21 +2.96 24...fxg5 25.Rd1 Rc8 26.Qd6+ Kh5
27.g4+ Kxg4 28.Qd4+ Kh5 29.Qxd7 Qxd7
30.Rxd7 Rc1+ 31.Kg2 Rc2+ 32.Kf3 Rxb2
33.Rxg7 Kh6 34.Re7 Rxh2 35.Rxb7 Rxa2 (33.359.237) 406

16.01 3:21 +2.82 24...fxg5 25.Rd1 Rc8 26.Qd6+ Kh5
27.g4+ Kxg4 28.h3+ Kh4 29.Qd4+ Kh5
30.Qxd7 Qxd7 31.Rxd7 Rc1+ 32.Kg2 Rc2+
33.Kf3 Rxb2 34.Rxg7 Kh6 35.Rd7 Kg6
36.Ke3 (82.220.505) 407

17.01 11:05 +3.72 24...fxg5 25.Rd1 Rc8 26.Qd6+ Kh5
27.g4+ Kxg4 28.h3+ Kh4 29.Qd4+ Kh5
30.Qxd7 Qxd7 31.Rxd7 Rc1+ 32.Kg2 Rc2+
33.Kf3 Rxb2 34.Rxg7 Kh6 35.Rc7 Rxa2
36.Ke4 b5 (267.059.522) 401

18.01 24:55 +3.76 24...fxg5 25.Rd1 Rc8 26.Qd6+ Kh5
27.g4+ Kxg4 28.h3+ Kh4 29.Qd4+ Kh5
30.Qxd7 Qxd7 31.Rxd7 Rc1+ 32.Kg2 Rc2+
33.Kf3 Rxb2 34.Rxg7 Kh6 35.Rc7 Rxa2
36.Rxb7 a5 37.Ke4 (600.163.235) 401

19.01 51:17 +2.92 24...fxg5 25.Rd1 Rc8 26.Qd3+ Kh6
27.h4 Rc7 28.hxg5+ Kxg5 29.Qd2+ Kg6
30.Qf4 Rc6 31.Qd4 Rb6 32.Qd3+ Kh6
33.Qd2+ g5 34.Qxd7 Qxd7 35.Rxd7 Rxb2
36.Kh2 Kg6 37.Kg3 (1.246.541.590) 405

20.01 135:47 +1.68 24...fxg5 25.Rd1 Rc8 26.Qd3+ Kh6
27.h4 Rc6 28.Qxd7 Qxd7 29.hxg5+ Kxg5
30.Rxd7 Rb6 31.b3 Kf6 32.Kh2 Ra6
33.Rxb7 Rxa2 34.Kg3 a5 35.Ra7 Ra1
36.Kf4 Rf1+ 37.Ke4 Re1+ (3.133.909.686) 384

21.01 334:58 +1.68 24...fxg5 25.Rd1 Rc8 26.Qd3+ Kh6
27.h4 Rc6 28.Qxd7 Qxd7 29.hxg5+ Kxg5
30.Rxd7 Rb6 31.b3 Kf6 32.Kh2 Ra6
33.Rxb7 Rxa2 34.Kg3 a5 35.Ra7 Ra1
36.Kf4 Rf1+ 37.Ke4 Re1+ (8.429.017.079) 419


26.. fxg5 is probably a draw, but I don't find 24... Bc6 very convincing as a winning attempt, after 25. Qxd8 Rxd8 26. Be3, is Black's extra pawn really much of an asset? Did Zukertort not just blunder towards the end? A short analysis of the position after 24... Bc6 25. Qxd8 Rxd8 26. Be3 gives :

[D]3r4/pp4pp/2b2pk1/8/8/4B3/PP4PP/4R2K b - -

Engine: Ancalagon 1.08 Testversion 002 (256 MB)
by Tord Romstad

9.00 0:00 +1.50 26...Bd5 27.b3 b6 28.Kg1 Kf5 29.Kf2 Be4
30.Rc1 Ke5 (62.067) 795

10.00 0:00 +1.49 26...Bd5 27.b3 b6 28.Kg1 Kf5 29.Kf2 Be4
30.Kg3 Rd3 31.Kf2 (103.130) 825

11.00 0:00 +1.45 26...Bd5 27.b3 b6 28.Kg1 Kf5 29.Kf2 Be4
30.g3 Ke5 31.Bf4+ Kf5 32.Ke3 Rd3+
33.Ke2 (186.024) 794

12.00 0:00 +1.45 26...Bd5 27.b3 b6 28.Kg1 Kf5 29.Kf2 Be4
30.g3 Ke5 31.Bf4+ Kf5 32.Ke3 Rd3+
33.Ke2 (335.771) 795

13.00 0:00 +1.41 26...Bd5 27.b3 b6 28.Kg1 Kf5 29.Kf2 Be4
30.g3 Ke5 31.Bf4+ Kf5 32.Ke3 Rd3+
33.Ke2 a6 (615.615) 758

14.01 0:03 +1.37 26...Bd5 27.b3 b6 28.Kg1 Kf5 29.Kf2 Be4
30.Rc1 Rd7 31.g3 Re7 32.Ke2 Ke5
33.Bf4+ Kd4 34.Be3+ Kd5 35.Rd1+ Ke5 (2.360.425) 715

15.01 0:16 +1.35 26...Bd5 27.a3 b6 28.Kg1 Kf5 29.Rd1 Ke6
30.Re1 Kd7 31.Kf2 Re8 32.Rd1 Kd6
33.Bd4 Rc8 (10.503.801) 644

15.02 0:24 +1.37 26...Kf5 27.Kg1 a6 28.Kf2 Be4 29.Rc1 Rd7
30.Rc8 Ke6 31.Bf4 Bd5 32.a3 Kf5
33.Be3 Ke5 (15.227.565) 629

16.01 0:42 +1.37 26...Kf5 27.Kg1 a6 28.Kf2 Be4 29.Rc1 Rd7
30.Rc8 Ke6 31.Bf4 Bd5 32.a3 Kf5
33.Rc7 Rd8 34.Ke3 Re8+ 35.Kd4 Kxf4
36.Kxd5 (27.082.270) 642

17.01 1:27 +1.33 26...Kf5 27.Kg1 a6 28.Kf2 Bd5 29.Rd1 Ke6
30.Bf4 Rc8 31.a3 Rc2+ 32.Rd2 Rc4
33.g3 b6 34.Ke3 Re4+ 35.Kd3 Kf5 (57.074.258) 650

17.04 2:01 +1.35 26...Bd5 27.a3 a6 28.Kg1 Be4 29.Kf2 Rd7
30.Rc1 f5 31.h3 Kf6 32.g3 Ke6 33.Bf4 Rd3
34.Rc7 Kf6 (78.825.160) 650

18.01 3:15 +1.33 26...Bd5 27.a3 a6 28.Kg1 Be4 29.Kf2 Rd7
30.Re2 Kf5 31.Rd2 Bd5 32.Bd4 Ke4
33.Bb6 Kf4 34.b3 Kf5 35.Bc5 (127.881.356) 653

19.01 6:26 +1.33 26...Bd5 27.a3 a6 28.Kg1 Be4 29.Kf2 Kf5
30.Rc1 g6 31.Rc7 h5 32.g3 Ke6 33.Ke2 f5
34.Bg5 Re8 35.Ke3 b6 (252.935.297) 653

20.01 12:33 +1.31 26...Bd5 27.a3 a6 28.Kg1 Be4 29.Kf2 Kf5
30.Rc1 g6 31.Rc7 h5 32.g3 Ke6 33.Ke2 f5
34.Bg5 Re8 35.Ke3 b6 36.Kd4 (494.829.196) 656

21.01 36:10 +1.31 26...Bd5 27.a3 a6 28.Kg1 Be4 29.Kf2 Kf5
30.g4+ Ke5 31.Rc1 Rd7 32.Kg3 f5
33.Rc5+ Kf6 34.gxf5 Bxf5 35.Rc1 Ke6
36.Kf4 Bd3 37.Bc5 Rf7+ 38.Ke3 (1.399.029.682) 644

best move: Bc6-d5 time: 37:03.125 min n/s: 643.998 nodes: 1.431.690.000


Trying to see how I can let Glaurung generate checks at depths deeper than 0 in the QSearch, -i.e. at negative depths, smaller depths than Depth(0)- but probably I'm missing the whole point of how cleverly that is done in Glaurung, with checks in QSearch at Depth(-1) the results I got were exactly the same as with Depth(0), I did not expect that and understand only a very little bit why that is so, checks generated at >= Depth (-2) does at least give different evals and nodecounts but not such a good analysis, I'm guessing Depth(-3) would be the same as Depth (-2), checks generated at depths >= Depth (-4) is what I tried in this analysis, it is not too bad but I'm not sure it would help finding any eternal checks which is what I was mostly after as that seems the reason 24... fxg5 is a draw.

At least Ancalagon with this change or with normal QSearch manages to find 26. h4 which I liked a bit more as a bit easier testposition:

[D]2rq4/pp1b2pp/6k1/3Q2p1/8/8/PP4PP/3R3K w - -

Engine: Ancalagon 1.08 Testversion 002 (256 MB)
by Tord Romstad

8.00 0:00 -3.01 26.Qd6+ Kh5 27.g4+ Kxg4 28.Qd4+ Kh5
29.Qxd7 Qxd7 30.Rxd7 Rc1+ 31.Kg2 Rc2+
32.Kf3 g4+ 33.Ke3 Rxb2 34.Rxg7 (315.996) 348

9.01 0:01 -3.01 26.Qd6+ Kh5 27.g4+ Kxg4 28.Qd4+ Kh5
29.Qxd7 Qxd7 30.Rxd7 Rc1+ 31.Kg2 Rc2+
32.Kf3 g4+ 33.Ke3 Rxb2 34.Rxg7 (594.959) 388

10.01 0:02 -3.01 26.Qd6+ Kh5 27.g4+ Kxg4 28.Qd4+ Kh5
29.Qxd7 Qxd7 30.Rxd7 Rc1+ 31.Kg2 Rc2+
32.Kf3 g4+ 33.Ke3 Rxb2 34.Rxg7 (1.147.266) 421

11.01 0:05 -2.96 26.Qd6+ Kh5 27.g4+ Kxg4 28.Qd4+ Kh5
29.Qxd7 Qxd7 30.Rxd7 Rc1+ 31.Kg2 Rc2+
32.Kf3 Rxb2 33.Rxg7 Kh6 34.Rd7 Rxh2
35.Rxb7 Rxa2 (2.357.806) 437

12.01 0:10 -2.98 26.Qd6+ Kh5 27.g4+ Kxg4 28.Qd4+ Kh5
29.Qxd7 Qxd7 30.Rxd7 Rc1+ 31.Kg2 Rc2+
32.Kf3 Rxb2 33.Rxg7 h6 34.h3 Rb4 (4.655.564) 441

13.01 0:23 -2.82 26.Qd6+ Kh5 27.g4+ Kxg4 28.h3+ Kh4
29.Qd4+ Kh5 30.Qxd7 Qxd7 31.Rxd7 Rc1+
32.Kg2 Rc2+ 33.Kf3 Rxb2 34.Rxg7 Kh6
35.Rd7 Kg6 (10.638.623) 446

14.01 0:48 -2.82 26.Qd6+ Kh5 27.g4+ Kxg4 28.h3+ Kh4
29.Qd4+ Kh5 30.Qxd7 Qxd7 31.Rxd7 Rc1+
32.Kg2 Rc2+ 33.Kf3 Rxb2 34.Rxg7 Kh6
35.Rd7 Kg6 36.Ke3 (21.953.582) 451

15.01 1:51 -3.76 26.Qd6+ Kh5 27.g4+ Kxg4 28.h3+ Kh4
29.Qd4+ Kh5 30.Qxd7 Qxd7 31.Rxd7 Rc1+
32.Kg2 Rc2+ 33.Kf3 Rxb2 34.Rxg7 Kh6
35.Rd7 Rxa2 36.Rxb7 a5 37.Ke4 (50.680.968) 453

15.02 3:10 -2.74 26.h4 Rc7 27.Qd3+ Kh6 28.hxg5+ Kxg5
29.Qd2+ Kg6 30.Qf4 h6 31.Qe4+ Bf5
32.Qxf5+ Kxf5 33.Rxd8 Rc2 34.Rd5+ Ke4
35.Rb5 Rc7 36.Kh2 g6 (85.024.272) 445

16.01 11:21 -1.78 26.h4 Rc6 27.hxg5 Qxg5 28.Qxg5+ Kxg5
29.Rxd7 Rb6 30.Rxg7+ Kh6 31.Rg4 Rxb2
32.Ra4 a6 33.Kh2 Kg5 34.Kg3 Rc2
35.Kf3 (299.975.081) 439

17.01 16:14 -1.78 26.h4 Rc6 27.hxg5 Qxg5 28.Qxg5+ Kxg5
29.Rxd7 Rb6 30.Rxg7+ Kh6 31.Rg4 Rxb2
32.Ra4 a6 33.Kh2 Kg5 34.Kg3 Re2
35.Ra5+ Kf6 36.Kf3 (434.961.670) 446

18.01 28:06 -1.72 26.h4 Rc6 27.hxg5 Qxg5 28.Qxg5+ Kxg5
29.Rxd7 Rb6 30.Rxg7+ Kh6 31.Rg4 Rxb2
32.Ra4 a6 33.Kh2 Kg5 34.Kg3 Rc2
35.Kf3 Kf5 36.Rf4+ Ke5 37.Re4+ Kd5 (761.232.562) 451

best move: h2-h4 time: 32:40.422 min n/s: 451.317 nodes: 884.730.000

Eelco

[Dr.Ex]

26.. fxg5 is probably a draw, but I don't find 24... Bc6 very convincing as a winning attempt, after 25. Qxd8 Rxd8 26. Be3, is Black's extra pawn really much of an asset? Did Zukertort not just blunder towards the end? A short analysis of the position after 24... Bc6 25. Qxd8 Rxd8 26. Be3 gives :


I find it very convincing. The main line is a perpetual. The rook endings with double pawn up for black are an easy draw for white.
So why not try your luck in an opposite coloured bishop ending with rooks still on the board? In these endings with rooks still on the board the stronger side almost always has some winning chances.