The horizon effect

[Eelco de Groot]

Rybka v 2.2 found it in 1 sec.

I don't know if that is correct Gerold, here Rybka 2.2n2 32 bit does not see anything in the first half hour...


[D]8/3P3k/n2K3p/2p3n1/1b4N1/2p1p1P1/8/3B4 w - -

Engine: Rybka 2.2n2 mp 32-bit PVtips5menbasesSO (256 MB)
by Vasik Rajlich

2.00 0:00 -3.92 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kf5
4.Bc2+ (1.800) 38

2.00 0:00 -2.02 1.Nxe3 Nf7+ 2.Ke7 Kg8 3.Bc2 c4+
4.Kf6 (2.634) 42
.
.
.

23.01 17:12 -1.60 1.d8Q Nf7+ 2.Ke7 Nxd8 3.Kxd8 e2
4.Bxe2 c4 5.Bxc4 Ba5+ 6.Ke7 Nc5
7.Ne3 Bc7 8.Ba2 Kg6 9.Bb1+ Kg5
10.Nd1 Be5 11.Bc2 Kh5 12.Nf2 Na6
13.g4+ Kg5 14.Nd3 (116.555.548) 115

24.01 30:09 -1.64 1.d8Q Nf7+ 2.Ke7 Nxd8 3.Kxd8 e2
4.Bxe2 c4 5.Bxc4 Ba5+ 6.Ke7 Nc5
7.Ne3 Bc7 8.Ba2 Kg6 9.Bb1+ Kg5
10.Nd1 Be5 11.Bc2 Kh5 12.Nf2 Na6
13.Nd3 (212.516.483) 120

best move: d7-d8Q time: 38:11.828 min n/s: 288.395.849 nodes: 288.395.849


Rainbow Serpent also does not see anything but at least it plays the correct Nf6+ and thinks a perpetual check is possible..

8/3P3k/n2K3p/2p3n1/1b4N1/2p1p1P1/8/3B4 w - -

Engine: Rainbow Serpent 1.8(s)dc Build 066 (Athlon 2009 MHz, 256 MB)
by Tord Romstad, Marco Costalba, Joona Kiiski

1.00 0:00 +4.48 1.Bc2+ (354) 1

2.00 0:00 -1.01 1.Bc2+ Kg7 2.Nxe3 Nf7+ 3.Ke7 (1.381) 4

2.00 0:00 -0.32 1.Nf6+ Kg6 2.Bc2+ Kxf6 3.d8Q+ Kf7
4.Ke5 (2.010) 6

3.00 0:00 +0.92 1.Nf6+ Kg6 2.Bc2+ Kxf6 3.d8Q+ Kg7
4.Qe7+ Nf7+ 5.Kd5 (8.701) 26

4.00 0:00 +0.92 1.Nf6+ Kg6 2.Bc2+ Kxf6 3.d8Q+ Kg7
4.Qe7+ Nf7+ 5.Kd5 (19.111) 55

5.00 0:00 +1.18 1.Nf6+ Kg6 (40.534) 108

6.00 0:00 +0.51-- 1.Nf6+ Kg7 (87.834) 200

6.00 0:00 +0.52-- 1.Nf6+ Kg7 (118.847) 245

6.00 0:00 +0.01-- 1.Nf6+ Kg7 (126.964) 253

6.00 0:00 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (146.473) 283

7.00 0:00 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (154.521) 291

8.00 0:00 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (427.057) 455

9.01 0:01 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (660.202) 515

10.01 0:01 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (990.978) 561

11.01 0:02 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (1.485.784) 597

12.01 0:02 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (1.737.367) 610

13.01 0:03 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (2.290.806) 637

14.01 0:04 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (2.938.711) 650

15.01 0:05 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (3.703.885) 664

16.01 0:07 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (4.903.730) 674

17.01 0:12 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (8.888.468) 685

18.01 0:21 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (14.840.223) 690

19.01 0:30 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (20.963.990) 696

20.01 0:45 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (31.871.375) 703

21.01 1:56 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (80.743.744) 690

22.01 3:15 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (133.515.683) 683

23.01 4:41 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (191.229.537) 678

24.01 10:55 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (443.357.593) 676

25.01 21:15 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (870.093.264) 681

26.01 42:24 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (1.772.357.760) 696

27.01 80:49 0.00 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kg7
4.Nh5+ (3.406.396.345) 702

best move: Ng4-f6 time: 131:33.516 min n/s: 717.641 nodes: 5.664.712.222

[bob]

[d] 8/3P3k/n2K3p/2p3n1/1b4N1/2p1p1P1/8/3B4 w - - 0 1
White to play and win.

Fruit cannot see the winning move due to the horizon effect - how about your engine?

This does not appear to be a "horizon effect" position. With a true horizon effect position, if you search a little deeper, you see the punch line. The point was that you played some sort of delaying move to push the "punch line" beyond the search horizon. This looks more like a zugzwang position.

[Don]

[d] 8/3P3k/n2K3p/2p3n1/1b4N1/2p1p1P1/8/3B4 w - - 0 1
White to play and win.

Fruit cannot see the winning move due to the horizon effect - how about your engine?

This does not appear to be a "horizon effect" position. With a true horizon effect position, if you search a little deeper, you see the punch line. The point was that you played some sort of delaying move to push the "punch line" beyond the search horizon. This looks more like a zugzwang position.

Or it may simply be a test of under-promotion.

[OliverUwira]

Isn't that strange? Without underpromotions the mate would show up earlier. Also zugzwang doesn't take part in the mating line.

Isn't it more likely that some White move gets pruned or reduced, so that it even takes a deeper search to discover it?

[Don]

Isn't that strange? Without underpromotions the mate would show up earlier. Also zugzwang doesn't take part in the mating line.

Isn't it more likely that some White move gets pruned or reduced, so that it even takes a deeper search to discover it?

I tried this on Komodo 1.0 which does not have under-promotion. It solves it more quickly that the version of Komodo I was using. So it is likely some kind of pruning issue.

When I say, "solve" what I mean is that it plays Nf6 with a positive score. Komodo likes Nf6 even very early but the score is negative. Komodo 1.0 on my machine see's 64 centipawns at depth 15 in 4.5 seconds which goes from losing at 14 ply to very good at 15 ply and the score gradually increases with depth after that.

[Eelco de Groot]

[d] 8/3P3k/n2K3p/2p3n1/1b4N1/2p1p1P1/8/3B4 w - - 0 1
White to play and win.

Fruit cannot see the winning move due to the horizon effect - how about your engine?

This does not appear to be a "horizon effect" position. With a true horizon effect position, if you search a little deeper, you see the punch line. The point was that you played some sort of delaying move to push the "punch line" beyond the search horizon. This looks more like a zugzwang position.

Or it may simply be a test of under-promotion.

I would think it is still mostly a horizon effect. Move 4. ... Nf7+ seems to refute the promotion because the resulting Queen is captured and Black has two more pawns to promote seemingly leaving a huge material plus. The underpromotions are not a problem why Rybka does not see it, it only can't see Bishop underpromotions. The knight underpromotions just make the line longer. I suspect the main difference why the clones see it is that BB once mentioned in the Rybka forum one of the few differences in eval between Rybka and the clones was that in the endgame the clones see that a King is locked up in a column, presumably only the A- or H- column and if you give "King Safety" points for that enough, or extend because of it, you look deeper into the promotion line. Because of the huge material deficit I think Stockfish heavily reduces the d8Q Nf7+ promotionline, cuts it very short. For instance after the main line sequence


[FEN "8/3P3k/n2K3p/2p3n1/1b4N1/2p1p1P1/8/3B4 w - -"]

1. Nf6+ Kg7 2. Nh5+ Kg6 3. Bc2+ Kxh5 *


8/3P4/n2K3p/2p3nk/1b6/2p1p1P1/2B5/8 w - -

Engine: Stockfish 1.8 JA (256 MB)
by Tord Romstad, Marco Costalba, Joona Kiiski

1.00 0:00 -0.96 4.Bd1+ (279)

2.00 0:00 -0.48 4.Bd1+ Kg6 5.Bc2+ Kg7 6.d8Q (415)

3.00 0:00 -6.74 4.Bd1+ Kg6 5.Bc2+ Kg7 6.Ke5 Nf7+
7.Ke4 e2 (655)

3.00 0:00 -2.30 4.g4+ Kxg4 5.Bd1+ Kf4 6.d8Q Nf7+ (792)

4.00 0:00 -8.28 4.g4+ Kxg4 5.Bf5+ Kxf5 6.d8Q Nf7+ (1.119)

4.00 0:00 -7.47 4.Bd1+ Kg6 5.Bc2+ Kg7 6.Bd1 Ba5 (2.018)

5.00 0:00 -7.39 4.Bd1+ Kg6 5.Kd5 Nf7 6.Be2 (4.794)

6.00 0:00 -7.83 4.Bd1+ Kg6 5.Ke5 Ba5 6.Kf4 Nb8
7.Bc2+ Kf6 8.d8N Bxd8 9.Kxe3 (11.741) 733

7.00 0:00 -8.32-- 4.Bd1+ Kg6 (17.554) 548

7.00 0:00 -8.80-- 4.Bd1+ Kg6 (27.426) 583

7.00 0:00 -8.84 4.Bd1+ Kg6 5.Kd5 Nf7 6.Kc6 Nb8+
7.Kc7 Nxd7 8.Kxd7 Kf5 9.Kc6 (43.601) 692

8.00 0:00 -9.05 4.Bd1+ Kg6 5.Kc6 Nb8+ 6.Kc7 Nxd7
7.Kxd7 Ne4 8.Bc2 Kf5 9.Kc6 (64.609) 687

9.00 0:00 -9.21 4.Bd1+ Kg6 5.Kd5 Nf7 6.Kc4 Ne5+
7.Kb3 Nxd7 8.Kc2 Ne5 9.Be2 Nc7 (88.257) 706

10.00 0:00 -9.49-- 4.Bd1+ Kg6 (121.498) 773

10.00 0:00 -9.77-- 4.Bd1+ Kg6 (164.602) 810

10.00 0:00 -9.77 4.Bd1+ Kg6 5.Kd5 Nf7 6.Kc4 Ne5+
7.Kb3 Nxd7 8.Bc2+ Kg5 9.Bd1 Ne5
10.Be2 Nc7 11.Kc2 Nd5 (190.352) 810

11.00 0:00 -10.42-- 4.Bd1+ Kg6 (277.636) 807

11.00 0:00 -10.54 4.Bd1+ Kg6 5.Kd5 Nf7 6.Kc4 Ne5+
7.Kb3 Nxd7 8.Kc2 Ne5 9.Be2 Nc7
10.Kc1 Ne6 11.Kd1 Kf5 (379.402) 837

12.00 0:00 -10.86 4.Bd1+ Kg6 5.Kd5 Nf7 6.g4 Nb8 7.Kc4 Ne5+
8.Kb3 Nbxd7 9.Kc2 Kg5 10.Be2 Kf4
11.Bb5 c4 (700.926) 815

12.13 0:02 -10.50 4.Ke7 e2 5.d8Q e1Q+ 6.Kf6 Qe6+ 7.Kg7 Kg4
8.Qf6 Kxg3 9.Kg6 Qe2 10.Qd6+ Kg2
11.Qd5+ Kf2 12.Qf5+ Nf3 13.Kxh6 c4
14.Kh7 Qe7+ 15.Kh8 Qh4+ 16.Qh7 (1.850.300) 822

13.01 0:02 -10.94-- 4.Ke7 e2 (2.279.477) 819

13.01 0:03 -10.82 4.Ke7 e2 5.d8Q e1Q+ 6.Kf6 Qe6+ 7.Kg7 Kg4
8.Qf6 Kxg3 9.Ba4 Qd5 10.Qg6 Kf4
11.Kxh6 Ne6 (3.124.242) 816

14.01 0:04 -11.19-- 4.Ke7 e2 (3.417.364) 815

14.03 0:04 -10.86 4.Bd1+ Kg6 5.Kd5 Nf7 6.g4 (3.532.267) 816

15.01 0:04 -11.07-- 4.Bd1+ Kg6 (3.870.060) 820

15.01 0:05 -11.27-- 4.Bd1+ Kg6 (4.354.627) 822

15.01 0:06 -11.67-- 4.Bd1+ Kg6 (5.212.694) 827

15.01 0:10 -12.48-- 4.Bd1+ Kg6 (8.351.747) 815

15.01 0:13 -12.80 4.Bd1+ Kg6 5.Kd5 Nf7 6.Kc4 Ne5+
7.Kb3 Nxd7 8.Ka2 Ne5 9.g4 Nc7 10.g5 Kxg5
11.Kb1 Ne6 12.Kc1 Nd4 13.Ba4 (10.742.778) 812

15.02 0:13 -12.48 4.Bb3 e2 5.d8Q e1Q 6.Kc6 Qe4+ 7.Kb5 Qb7+
8.Kc4 Nc7 9.Bd1+ Kg6 10.Bc2+ Kf7
11.Qd7+ Kf6 12.Qf5+ Ke7 13.Kd3 Nce6
14.Bb3 Qg2 15.Qd5 Qxg3+ 16.Kc2 Nd4+
17.Kc1 (11.288.337) 811

16.01 0:16 -14.14-- 4.Bb3 e2 (13.706.203) 812

16.02 0:17 -12.80 4.Bd1+ Kg6 5.Kd5 Nf7 6.g4 Nb8 7.d8N Nxd8
8.Kc4 Kg5 9.Kd3 Kf4 10.g5 c4+ 11.Kc2 hxg5
12.Kc1 Nbc6 13.Bh5 g4 14.Kd1 (13.883.743) 813

17.01 0:19 -13.93-- 4.Bd1+ Kg6 (15.906.251) 819

17.01 0:25 -14.50 4.Bd1+ Kg6 5.Ke5 Ba5 6.Kf4 Nb4
7.Kxe3 c2 8.Bxc2+ Nxc2+ 9.Kd3 Nd4
10.Kc4 Nge6 11.Kd5 Kf6 12.Ke4 (21.146.535) 828

18.01 0:30 -14.94 4.Bd1+ Kg6 5.Ke5 Ba5 6.Kf4 Nb4
7.Kxe3 c2 8.Bxc2+ Nxc2+ 9.Kd3 Nd4
10.Kc4 Nge6 11.Kd5 Kf6 12.g4 (24.986.498) 826

19.01 0:39 -15.68 4.Bd1+ Kg6 5.Kd5 Nf7 6.g4 Nb8 7.Kc4 Ne5+
8.Kb3 Nbxd7 9.Kc2 Nc6 10.Kd3 Nd4
11.Kxe3 c2 12.Bxc2+ Nxc2+ 13.Kd3 Nd4
14.Ke3 Kg5 15.Ke4 Kxg4 (32.885.267) 830

20.01 0:49 -15.73 4.Bd1+ Kg6 5.Kd5 Nf7 6.Kc4 Ne5+
7.Kb3 Nxd7 8.g4 Ne5 9.Kc2 Nc6 10.Kd3 Nd4
11.Kxe3 c2 12.Bxc2+ Nxc2+ 13.Ke2 Kg5
14.Kd1 Ne3+ 15.Ke2 Kf4 16.Kd3 Nc7
17.g5 (40.923.697) 833

21.01 0:59 -16.18-- 4.Bd1+ Kg6 (49.756.371) 834

21.13 1:07 -15.29++ 4.d8Q Nf7+ (56.177.948) 836

21.13 1:07 -14.84++ 4.d8Q Nf7+ (56.300.120) 836

21.13 1:07 -13.95++ 4.d8Q Nf7+ (56.428.043) 836

21.13 1:07 -12.18++ 4.d8Q Nf7+ (56.681.921) 837

21.13 1:10 -8.62++ 4.d8Q Nf7+ (58.913.364) 835

21.13 1:17 -1.51++ 4.d8Q Nf7+ (64.942.088) 839

21.13 2:38 +2.78 4.d8Q Kg4 5.Qe8 Kf3 6.Qh5+ Kxg3
7.Qe2 c4+ 8.Kd5 Bc5 9.Bd1 c2 10.Bxc2 Bb6
11.Kxc4 Nc7 12.Bd1 Nge6 13.Qg4+ Kf2
14.Qh4+ Kg2 15.Qxh6 Kg3 (141.020.854) 890

22.01 3:38 +4.44 4.d8Q Kg4 5.Qe8 Kf3 6.Qh5+ Kxg3
7.Qe2 c4+ 8.Kd5 Bc5 9.Bd1 c2 10.Bxc2 Bb6
11.Kxc4 Nc7 12.Qe1+ Kg2 13.Qb4 Na8
14.Ba4 Kf2 15.Qf8+ Ke1 16.Bc6 e2
17.Qxh6 (195.689.150) 895

23.01 4:31 +4.40 4.d8Q Kg4 5.Qe8 c4+ 6.Kc6 Ba3 7.Bd1+ Kh3
8.Qxe3 Nb4+ 9.Kb5 Nd3 10.Qe2 Kxg3
11.Qg4+ Kh2 12.Qh4+ Kg1 13.Qxc4 Nf2
14.Qxc3 Bd6 15.Bc2 Bf4 16.Qg7 Kg2
17.Qxh6 (242.864.657) 895

24.01 5:14 +4.60 4.d8Q Kg4 5.Qe8 c4+ 6.Kc6 Ba3 7.Bd1+ Kh3
8.Qxe3 Nb4+ 9.Kb5 Nd3 10.Qe2 Kxg3
11.Qg4+ Kh2 12.Qh4+ Kg1 13.Qxc4 Nf2
14.Qxc3 Bd6 15.Qf6 Nge4 16.Qg6+ Bg3
17.Bc2 (282.138.752) 897

25.01 5:51 +4.36-- 4.d8Q Kg4 (316.497.967) 900

25.01 6:03 +4.12-- 4.d8Q Kg4 (327.346.823) 901

25.01 7:31 +4.52 4.d8Q Kg4 5.Qe8 Kf3 6.Qh5+ Kxg3
7.Qe2 c4+ 8.Kd5 Bc5 9.Bd1 c2 10.Bxc2 Bb6
11.Kxc4 Nc7 12.Qe1+ Kg2 13.Qb4 e2
14.Qd2 Kf3 15.Bd3 e1N 16.Qe2+ Kf4
17.Qxe1 (405.192.316) 896

26.01 8:21 +4.44 4.d8Q Kg4 5.Qe8 Kf3 6.Qh5+ Kxg3
7.Qe2 c4+ 8.Kd5 Bc5 9.Bd1 c2 10.Bxc2 Ba7
11.Kxc4 Nc7 12.Qe1+ Kg2 13.Qa5 e2
14.Qe1 Kf3 15.Bd3 Kf4 16.Qxe2 Nce6
17.Qh2+ (449.315.200) 895

27.01 9:25 +4.56++ 4.d8Q Kg4 (507.724.596) 898

27.01 10:27 +4.68++ 4.d8Q Kg4 (564.615.768) 899

27.01 12:09 +4.76 4.d8Q Kg4 5.Qe8 Kf3 6.Qh5+ Kxg3
7.Qe2 c4+ 8.Kd5 Bc5 9.Bd1 c2 10.Bxc2 Ba7
11.Kxc4 Nc7 12.Qe1+ Kg2 13.Qa5 e2
14.Qe1 Kf3 15.Bd3 Kf4 16.Bxe2 Nce6 (657.937.285) 902

28.01 19:24 +4.84 4.d8Q Kg4 5.Qe8 c4+ 6.Kc6 Ba3 7.Bd1+ Kh3
8.Qxe3 Nb4+ 9.Kb5 Nd3 10.Qe2 Kxg3
11.Qg4+ Kh2 12.Qh4+ Kg2 13.Qxh6 Bc1
14.Qc6+ Kf2 15.Qxc4 Ke3 16.Bc2 Ne1
17.Qxc3+ (1.055.193.695) 906

29.01 25:28 +5.09++ 4.d8Q Kg4 (1.380.497.362) 903

29.01 32:33 +5.13 4.d8Q Kg4 5.Qe8 c4+ 6.Kc6 Ba3 7.Bd1+ Kh3
8.Qxe3 Nb4+ 9.Kb5 Nd3 10.Qe2 Kxg3
11.Qg4+ Kh2 12.Qh4+ Kg2 13.Qxh6 Bc1
14.Qc6+ Kh2 (1.782.068.183) 912

30.01 47:59 +5.09 4.d8Q Kg4 5.Qe8 c4+ 6.Kc6 Kf3 7.Bd1+ Kg2
8.Qxe3 Bc5 9.Qxc3 Bf2 10.Qxc4 Nc5
11.Qd5+ Kg1 12.Qd6 Nce6 13.Bg4 Nd4+
14.Kd7 Ne4 15.Qxh6 Nxg3 16.Qf4 Kf1
17.Bf3 (2.670.338.132) 927

31.01 75:57 +5.01 4.d8Q Kg4 5.Qe8 c4+ 6.Kc6 Ba3 7.Bd1+ Kh3
8.Qxe3 Nb4+ 9.Kb5 c2 10.g4+ Kg2
11.Qd2+ Kg3 12.Qc3+ Kh4 13.Bxc2 Nxc2
14.Qxc2 c3 15.Qxc3 Bd6 16.Qf6 Bg3
17.Qf5 (4.266.484.648) 936

32.01 93:38 +5.13++ 4.d8Q Kg4 (5.274.820.338) 938

32.01 105:42 +5.09 4.d8Q Kg4 5.Qe8 c4+ 6.Kc6 Ba3 7.Bd1+ Kh3
8.Qxe3 Nb4+ 9.Kb5 c2 10.g4+ Kg2
11.Qd2+ Kg3 12.Qc3+ Kh4 (5.979.429.490) 942

33.01 182:16 +5.21++ 4.d8Q Kg4 (10.332.630.453) 944

33.01 224:46 +5.33++ 4.d8Q Kg4 (12.735.827.073) 944

33.01 260:29 +5.57++ 4.d8Q Kg4 (14.782.378.433) 945

33.01 289:03 +6.06++ 4.d8Q Kg4 (16.420.366.070) 946

33.01 305:33 +5.77 4.d8Q Kg4 5.Qf6 Kxg3 6.Qe5+ Kf3
7.Bd1+ Kg2 8.Qxe3 c4+ 9.Kc6 Bc5
10.Qxc3 Bf2 11.Qxc4 Nc5 12.Qg4+ Kh2
13.Qh5+ Kg2 14.Qxh6 Nce6 15.Qh5 Nd4+
16.Kd5 Be3 17.Qg4+ (17.367.404.187) 947

34.01 390:34 +6.50++ 4.d8Q Kg4 (22.379.626.230) 954

34.01 541:29 +7.23++ 4.d8Q Kg4 (31.361.117.225) 965

best move: d7-d8Q time: 554:06.172 min n/s: 965.440 nodes: 32.097.214.330

There may be other effects in nullwindow searches going on, but material plays a big role in determining the ReductionMatrix futility reduction depth so I think that is the main reason. At least in the PV Stockfish can see the punchline now after getting a little help

Eelco

[Jack Lad]

Hi Eelco

Terry claims that Rybka 3 can't see Qxc4 in this study of mine.
[d] 7k/2p1p2P/1p2p1P1/pp3Pp1/Q1qP4/6K1/6P1/8 w
Do you agree?

How does your version of Toga II and Stockfish do?

I will try to see if I can get Houdini to work on it.

[tmokonen]

The latest version of Chessmaster finds it pretty quickly:

FEN: 7k/2p1p2P/1p2p1P1/pp3Pp1/Q1qP4/6K1/6P1/8 w - - 0 1

The King 3.50 x64 2 cpu:
3001 00:00 2.126 106.300 -0.72 Qa3 Qxd4 Kf3 exf5 Qxe7
3001 00:00 2.876 143.800 +0.24 Qa1 Qd3+ Kg4 Qxf5+ Kh5
4001 00:00 8.324 208.100 -0.03 Qa1 b4 Qb1 exf5
5001 00:00 19.143 382.860 -0.73 Qa1 b4 Kg4 exf5+ Kxg5 b3 Kxf5
5001 00:00 22.260 371.000 -0.70 Qd1 exf5 Qb1 Qc3+ Kh2 a4 Qxb5
6001 00:00 53.344 484.945 -0.44 Qd1 b4 fxe6 Qc3+ Kg4 b3 Kxg5
7001 00:00 169.456 584.331 -0.57 Qd1 Qd5 f6 exf6 Qh1 Qd6+ Kf3 Qa3+ Ke2
8002 00:00 295.620 642.652 -0.67 Qd1 Qd5 f6 exf6 Qh5 Qd6+ Kh3 Qd8 Qd1
9002 00:00 476.335 661.576 -1.14 Qd1 Qd5 f6 exf6 Qh5 Qd6+ Kf3 Qf4+ Ke2 Qe4+ Kf1 Qd3+ Ke1 Qc3+ Ke2
10002 00:02 1.769.009 701.987 -1.01 Qd1 Qc3+ Kg4 b4 Kh5 Qg3 Qh1 exf5 Kh6 Qd6 Kxg5 Qxd4 Kxf5
11003 00:05 4.058.524 710.774 -0.26 Qd1 Qc3+ Kh2 Qe3 Qh5 Qf4+ g3 Qf2+ Kh1 Qxd4 Qxg5 b4 Qxe7 Qe4+ Kg1 Qxf5
12003 00:10 7.614.200 732.839 -0.41 Qd1 Qc3+ Kg4 Qe3 Kh5 exf5 d5 Qe5 Qd3 b4 Qb3 Qc3 Qb1
12003 00:12 9.218.075 739.813 +3.42 Qxc4 bxc4 f6 exf6 d5 exd5 Kg4 Kg7 Kf5 d4 Ke6 Kxg6 h8Q d3 Qxf6+ Kh5 Qf7+ Kh6 Qxc7
13003 00:14 10.962.422 767.139 +4.13 Qxc4 bxc4 f6 exf6 d5 exd5 Kg4 c3 Kf5 c2 Ke6 c1Q Kf7 Qb1 g7+ Kxh7 g8Q+ Kh6 Qg7+ Kh5 Qh8+ Kg4 Qh3+ Kf4 Qf3+ Ke5 Qxf6+ Ke4 Qg6+ Ke3 Qxb1 c5
14004 00:18 14.610.171 803.197 +4.46 Qxc4 bxc4 f6 exf6 d5 exd5 Kg4 c3 Kf5 c2 Ke6 c1Q Kf7 Qb1 g7+ Kxh7 g8Q+ Kh6 Qg7+ Kh5 Qh8+ Kg4 Qh3+ Kf4 Qf3+ Ke5 Qxf6+ Ke4 Qg6+ Ke5 Qxb1 c5 Kg6

[Jack Lad]

12003 00:12 9.218.075 739.813 +3.42 Qxc4 bxc4 f6 exf6 d5 exd5 Kg4 Kg7 Kf5 d4 Ke6 Kxg6 h8Q d3 Qxf6+ Kh5 Qf7+ Kh6 Qxc7
13003 00:14 10.962.422 767.139 +4.13 Qxc4 bxc4 f6 exf6 d5 exd5 Kg4 c3 Kf5 c2 Ke6 c1Q Kf7 Qb1 g7+ Kxh7 g8Q+ Kh6 Qg7+ Kh5 Qh8+ Kg4 Qh3+ Kf4 Qf3+ Ke5 Qxf6+ Ke4 Qg6+ Ke3 Qxb1 c5
14004 00:18 14.610.171 803.197 +4.46 Qxc4 bxc4 f6 exf6 d5 exd5 Kg4 c3 Kf5 c2 Ke6 c1Q Kf7 Qb1 g7+ Kxh7 g8Q+ Kh6 Qg7+ Kh5 Qh8+ Kg4 Qh3+ Kf4 Qf3+ Ke5 Qxf6+ Ke4 Qg6+ Ke5 Qxb1 c5 Kg6

Wow - only 12 secs!

That is a new record which beats Spark by 2 secs!

My old version of Fruit took 30 mins to find it.

[Terry McCracken]

Hi Eelco

Terry claims that Rybka 3 can't see Qxc4 in this study of mine.
[d] 7k/2p1p2P/1p2p1P1/pp3Pp1/Q1qP4/6K1/6P1/8 w
Do you agree?

How does your version of Toga II and Stockfish do?

I will try to see if I can get Houdini to work on it.

If you play 1. Qxc4..bc 2. f6..ef 3. d5..ed, R4 sees 4. Kg4 within 15 sec. as winning.

So from the initial position it would either take a very long time or not at all.