I gave it 1 cpu and 1024 mbytes hash
at depth 23 it found mate in 46
I could expect at bigger depths to see only shorter mates or the same score but houdini shows at depth 24 mate in 48 at depth 25 mate in 65 at depth 26 mate in 75 at depth 27 mate in 80 and at depth 28 mate in 85(note that mate in more than 50 moves means a draw by the 50 move rule so I think that houdini should get draw score for these long mates
only later it can see shorter mates.
[d]8/8/8/3k4/8/8/8/5BBK w - - 18 1
Houdini_3_x64:
4/9 00:00 340 0 +6.47 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-e5
5/9 00:00 556 0 +6.51 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-e5 Kg3-f3
6/11 00:00 999 0 +6.59 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-e5 Kg3-f3 Ke5-d5 Kf3-f4
7/11 00:00 2,034 2,034,000 +6.56 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-f5 Kg3-f3 Kf5-e5 Bf1-d3
7/11 00:00 2,665 2,665,000 +6.57 Bg1-f2 Kd5-e4 Kh1-g2 Ke4-f4 Bf1-d3 Kf4-e5 Kg2-f3
8/13 00:00 4,102 2,051,000 +6.69 Bg1-f2 Kd5-e4 Kh1-g2 Ke4-f4 Bf1-d3 Kf4-e5 Kg2-f3 Ke5-d5 Bd3-e4+ Kd5-e5
9/13 00:00 6,798 2,266,000 +6.69 Bg1-f2 Kd5-e4 Kh1-g2 Ke4-f4 Bf1-d3 Kf4-e5 Kg2-f3 Ke5-d5 Bd3-e4+ Kd5-e5
9/15 00:00 9,922 2,480,000 +6.75 Bf1-d3 Kd5-e5 Kh1-g2 Ke5-d5 Kg2-f3 Kd5-e5 Bd3-e4 Ke5-e6 Bg1-d4
10/17 00:00 13,425 2,237,000 +6.80 Bf1-d3 Kd5-e5 Kh1-g2 Ke5-d5 Kg2-f3 Kd5-e5 Bd3-e4 Ke5-e6 Bg1-d4 Ke6-d6 Kf3-f4
10/17 00:00 18,754 2,083,000 +6.82 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-e5 Kg3-g4 Ke5-e6 Bf1-c4+ Ke6-d6 Kg4-f4 Kd6-c6 Kf4-e5
11/17 00:00 30,807 2,053,000 +6.92 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-e5 Bf1-g2 Ke5-f5 Bg1-d4 Kf5-e6 Kg3-f4 Ke6-e7 Kf4-e5 Ke7-d7
12/17 00:00 43,962 1,998,000 +6.93 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-e5 Bf1-g2 Ke5-f5 Bg1-d4 Kf5-e6 Kg3-f4 Ke6-e7 Kf4-e5 Ke7-d7 Bd4-e3
13/19 00:00 73,065 2,029,000 +6.97 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-e5 Bf1-g2 Ke5-f5 Bg1-d4 Kf5-e6 Kg3-f4 Ke6-e7 Kf4-e5 Ke7-d7 Bd4-e3 Kd7-e7 Bg2-e4
14/21 00:00 129,082 1,926,000 +7.05 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-e5 Bf1-g2 Ke5-f5 Bg1-d4 Kf5-e6 Kg3-f4 Ke6-e7 Kf4-e5 Ke7-d7 Bd4-e3 Kd7-e7 Bg2-e4 Ke7-d7 Be4-d5
15/25 00:00 233,241 1,911,000 +7.17 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-e5 Bf1-g2 Ke5-f5 Bg1-d4 Kf5-e6 Kg3-f4 Ke6-e7 Kf4-e5 Ke7-d7 Bd4-c5 Kd7-c7 Bg2-d5 Kc7-d7 Bc5-d6 Kd7-d8
16/28 00:00 385,405 1,966,000 +7.19 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-e5 Bf1-g2 Ke5-f5 Bg1-d4 Kf5-e6 Kg3-f4 Ke6-e7 Kf4-e5 Ke7-f7 Bg2-d5+ Kf7-g6 Bd4-e3 Kg6-h5 Ke5-f5 Kh5-h4 Be3-f2+ Kh4-h5 Bd5-f3+ Kh5-h6 Bf2-d4
17/29 00:00 634,041 2,006,000 +7.22 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-e5 Bf1-g2 Ke5-f5 Bg1-d4 Kf5-e6 Kg3-f4 Ke6-e7 Kf4-e5 Ke7-f7 Bg2-d5+ Kf7-g6 Bd4-e3 Kg6-g7 Ke5-f5 Kg7-f8 Be3-c5+ Kf8-e8 Kf5-e5 Ke8-d8 Ke5-e6 Kd8-c7 Bc5-d4 Kc7-d8 Bd5-g2 Kd8-c8 Ke6-d6 Kc8-d8 Kd6-e6 Kd8-c8
18/32 00:00 981,761 2,053,000 +7.22 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-e5 Bf1-g2 Ke5-f5 Bg1-d4 Kf5-e6 Kg3-f4 Ke6-e7 Kf4-e5 Ke7-f7 Bg2-d5+ Kf7-g6 Bd4-e3 Kg6-g7 Ke5-f5 Kg7-f8 Be3-c5+ Kf8-e8 Kf5-e5 Ke8-d8 Ke5-e6 Kd8-c7 Bc5-d4 Kc7-d8 Bd5-g2 Kd8-c8 Ke6-d6 Kc8-d8 Kd6-e6 Kd8-c8
19/35 00:01 1,745,408 2,082,000 +7.27 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-e5 Bf1-g2 Ke5-f5 Bg1-d4 Kf5-e6 Kg3-f4 Ke6-e7 Bd4-c5+ Ke7-f6 Bg2-d5 Kf6-g6 Bc5-d4 Kg6-h6 Kf4-f5 Kh6-h5 Bd4-e3 Kh5-h4 Be3-f2+ Kh4-h3 Bd5-e4 Kh3-h2 Kf5-e5 Kh2-h3 Be4-f3 Kh3-h2
20/35 00:01 2,911,863 2,119,000 +7.33 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-e5 Bf1-d3 Ke5-e6 Kg3-f4 Ke6-d5 Kf4-f5 Kd5-d6 Bd3-e4 Kd6-d7 Bg1-c5 Kd7-c8 Kf5-e6 Kc8-c7 Bc5-d4 Kc7-d8 Bd4-b6+ Kd8-e8 Be4-c6+ Ke8-f8 Bb6-d4 Kf8-g8 Bc6-e4 Kg8-f8 Be4-f3 Kf8-e8 Bf3-e4 Ke8-d8 Bd4-b6+
21/35 00:02 4,553,779 2,144,000 +7.33 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-e5 Bf1-d3 Ke5-e6 Kg3-f4 Ke6-d7 Kf4-e5 Kd7-c7 Bd3-e4 Kc7-d7 Be4-d5 Kd7-c7 Ke5-e6 Kc7-d8 Bd5-e4 Kd8-e8 Bg1-d4 Ke8-d8 Bd4-b6+ Kd8-c8 Bb6-c5 Kc8-c7 Bc5-d4 Kc7-d8 Bd4-b6+
22/35 00:03 7,133,373 2,174,000 +7.33 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-e5 Bf1-d3 Ke5-e6 Kg3-f4 Ke6-d7 Kf4-e5 Kd7-c7 Bd3-e4 Kc7-d7 Be4-d5 Kd7-c7 Ke5-e6 Kc7-d8 Bd5-g2 Kd8-e8 Bg1-d4 Ke8-f8 Bg2-f3 Kf8-e8 Bf3-e4 Ke8-f8 Be4-f3 Kf8-e8
23/39+ 00:03 8,315,385 2,192,000 +7.63 Kh1-g2
23/39+ 00:03 8,344,304 2,193,000 +8.25 Kh1-g2
23/47+ 00:04 8,571,901 2,188,000 +15.04 Kh1-g2
23/47 00:04 10,615,273 2,263,000 +M46 Kh1-g2 Kd5-e6 Kg2-f3 Ke6-e5 Bf1-c4 Ke5-f5 Bg1-d4 Kf5-g5 Bc4-d3 Kg5-h4 Kf3-f4 Kh4-h5 Bd3-e2+ Kh5-g6 Be2-f3 Kg6-f7 Bf3-d5+ Kf7-g6 Bd5-e4+ Kg6-h6 Kf4-f5 Kh6-h7 Kf5-g4+ Kh7-g8 Kg4-f4 Kg8-f7 Be4-d5+ Kf7-g6
24/47- 00:05 11,656,911 2,296,000 +M48 Kh1-g2 Kd5-e6
24/47 00:05 12,578,648 2,328,000 +M48 Kh1-g2 Kd5-e6 Kg2-f3 Ke6-f5 Bf1-d3+ Kf5-e6 Kf3-f4 Ke6-f6 Bd3-c4 Kf6-e7 Kf4-e5 Ke7-f8 Bg1-d4 Kf8-g7 Ke5-f4+ Kg7-g6 Bc4-f1 Kg6-f7 Kf4-f5 Kf7-e7 Bf1-d3 Ke7-f7 Bd3-c4+ Kf7-f8 Kf5-e5 Kf8-g7
25/47- 00:05 13,528,768 2,348,000 +M50 Kh1-g2 Kd5-e6
25/47 00:07 17,460,279 2,425,000 +M65 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-e5 Bf1-d3 Ke5-d5 Kg3-f4 Kd5-c6 Kf4-f5 Kc6-d5 Bd3-e2 Kd5-c6 Be2-c4 Kc6-b7 Bc4-d3 Kb7-c6 Bd3-c4 Kc6-b7
26/47- 00:08 19,787,201 2,458,000 +M67 Kh1-g2 Kd5-e4
26/47 00:10 25,550,371 2,514,000 +M75 Kh1-g2 Kd5-e6 Kg2-f3 Ke6-e7 Kf3-f4 Ke7-e6 Bf1-g2 Ke6-f6 Bg1-d4+ Kf6-e6 Kf4-e4 Ke6-d7 Ke4-e5 Kd7-d8 Ke5-d6 Kd8-e8 Kd6-e6 Ke8-d8 Bg2-f1 Kd8-c7 Bf1-a6 Kc7-b8 Ba6-d3 Kb8-a8 Bd3-e4+ Ka8-b8 Be4-d5 Kb8-c7 Bd5-g2 Kc7-d8 Bg2-f1
27/47- 00:11 28,771,607 2,535,000 +M77 Kh1-g2 Kd5-e6
27/47 00:14 36,387,719 2,571,000 +M80 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-f5 Bf1-g2 Kf5-e5 Kg3-g4 Ke5-f6 Bg1-d4+ Kf6-e6 Kg4-f4 Ke6-d7 Kf4-f5 Kd7-c8 Kf5-e6 Kc8-b8 Bg2-e4 Kb8-c7 Bd4-g1 Kc7-d8 Bg1-b6+ Kd8-c8 Bb6-c5 Kc8-c7 Bc5-d4 Kc7-b8 Bd4-b6 Kb8-c8 Bb6-c5
28/47- 00:15 40,882,863 2,588,000 +M82 Kh1-g2 Kd5-e4
28/47 00:18 47,278,982 2,609,000 +M85 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-f5 Bf1-g2 Kf5-e6 Kg3-f4 Ke6-d7 Kf4-e5 Kd7-e7 Bg2-e4 Ke7-d7 Bg1-d4 Kd7-d8 Ke5-d6 Kd8-e8 Kd6-e6 Ke8-d8 Bd4-b6+ Kd8-c8 Be4-c6 Kc8-b8 Bc6-d5 Kb8-c8 Bd5-g2 Kc8-b8 Bb6-d4 Kb8-c7 Bd4-g1 Kc7-c8 Bg1-b6 Kc8-b8
29/47 00:20 52,931,360 2,625,000 +M23 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-f5 Bf1-g2 Kf5-e5 Kg3-g4 Ke5-f6 Bg1-d4+ Kf6-e6 Bg2-e4 Ke6-e7 Kg4-f5 Ke7-d6 Kf5-f6 Kd6-d7 Bd4-c5 Kd7-d8 Kf6-e5 Kd8-c7 Ke5-e6 Kc7-c8 Ke6-d6 Kc8-d8 Bc5-d4 Kd8-e8 Be4-d5 Ke8-f8 Kd6-e6 Kf8-e8 Bd4-b6 Ke8-f8 Ke6-f6 Kf8-e8
30/47- 00:23 61,159,951 2,640,000 +M25 Kh1-g2 Kd5-e4
30/47 00:24 64,627,068 2,648,000 +M25 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-e5 Bf1-g2 Ke5-f5 Bg1-d4 Kf5-e6 Kg3-f4 Ke6-e7 Kf4-e5 Ke7-f7 Bg2-d5+ Kf7-e7 Bd4-c5+ Ke7-d8 Ke5-d6 Kd8-e8 Bc5-d4 Ke8-f8 Kd6-e6 Kf8-e8 Bd5-c6+ Ke8-d8 Ke6-d6 Kd8-c8 Bc6-e4 Kc8-d8 Be4-g2 Kd8-e8 Bg2-d5 Ke8-f8
31/47 00:27 71,617,068 2,652,000 +M23 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-f5 Bf1-g2 Kf5-e5 Kg3-g4 Ke5-f6 Bg1-d4+ Kf6-e6 Bg2-e4 Ke6-e7 Kg4-f5 Ke7-d6 Kf5-f6 Kd6-d7 Bd4-c5 Kd7-c8 Kf6-e6 Kc8-c7 Bc5-d4 Kc7-c8 Ke6-d6 Kc8-d8 Be4-g2 Kd8-e8 Bg2-d5 Ke8-f8 Kd6-e6 Kf8-e8 Bd5-c6+ Ke8-d8 Ke6-d6 Kd8-c8
32/47- 00:37 100,205,627 2,671,000 +M25 Kh1-g2 Kd5-e4
32/47 00:40 108,771,228 2,670,000 +M24 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-e5 Bf1-g2 Ke5-f5 Bg1-d4 Kf5-e6 Kg3-f4 Ke6-d6 Kf4-f5 Kd6-e7 Bg2-e4 Ke7-d6 Be4-g2 Kd6-e7
33/47 00:46 123,395,298 2,657,000 +M21 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-f5 Bf1-g2 Kf5-e5 Kg3-g4 Ke5-f6 Kg4-f4 Kf6-e7 Kf4-f5 Ke7-d6 Bg1-d4 Kd6-e7 Bd4-c5+ Ke7-d7 Bg2-e4 Kd7-d8 Kf5-e6 Kd8-c7 Be4-d5 Kc7-b8 Bc5-b6 Kb8-c8 Bd5-g2 Kc8-b8 Bg2-e4 Kb8-c8 Ke6-d6 Kc8-b8 Kd6-c6 Kb8-c8 Be4-f5+ Kc8-b8
34/47 00:56 148,703,533 2,651,000 +M21 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-f5 Bf1-g2 Kf5-e5 Kg3-g4 Ke5-f6 Kg4-f4 Kf6-e7 Kf4-f5 Ke7-d6 Bg1-d4 Kd6-e7 Bd4-c5+ Ke7-d7 Bg2-e4 Kd7-d8 Kf5-e6 Kd8-c7 Be4-d5 Kc7-b8 Bc5-b6 Kb8-c8 Bd5-g2 Kc8-b8 Bg2-e4 Kb8-c8 Ke6-d6 Kc8-b8 Kd6-c6 Kb8-c8 Be4-f5+ Kc8-b8
35/47 01:08 181,413,060 2,656,000 +M18 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-f5 Kg3-f3 Kf5-f6 Kf3-f4 Kf6-e6 Bf1-c4+ Ke6-d6 Kf4-f5 Kd6-d7 Kf5-e5 Kd7-e7 Bg1-c5+ Ke7-d7 Bc4-b5+ Kd7-c7 Ke5-e6 Kc7-c8 Ke6-d6 Kc8-b7 Kd6-d7 Kb7-a8 Bb5-a6 Ka8-b8 Bc5-d4 Kb8-a8 Kd7-c6 Ka8-b8 Kc6-b6 Kb8-a8 Ba6-b7+ Ka8-b8 Bd4-e5+
36/47 01:17 205,385,919 2,653,000 +M18 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-f5 Kg3-f3 Kf5-f6 Kf3-f4 Kf6-e6 Bf1-c4+ Ke6-d6 Kf4-f5 Kd6-d7 Kf5-e5 Kd7-e7 Bg1-c5+ Ke7-d7 Bc4-b5+ Kd7-c7 Ke5-e6 Kc7-c8 Ke6-d6 Kc8-b7 Kd6-d7 Kb7-a8 Bb5-a6 Ka8-b8 Bc5-d4 Kb8-a8 Kd7-c6 Ka8-b8 Kc6-b6 Kb8-a8 Ba6-b7+ Ka8-b8 Bd4-e5+
37/47 01:31 240,394,480 2,645,000 +M18 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-f5 Kg3-f3 Kf5-f6 Kf3-f4 Kf6-e6 Bf1-c4+ Ke6-d6 Kf4-f5 Kd6-d7 Kf5-e5 Kd7-e7 Bg1-c5+ Ke7-d7 Bc4-b5+ Kd7-c7 Ke5-e6 Kc7-c8 Ke6-d6 Kc8-b7 Kd6-d7 Kb7-b8 Bb5-a6 Kb8-a8 Bc5-d4 Ka8-b8 Kd7-c6 Kb8-a8 Kc6-b6 Ka8-b8 Bd4-e5+ Kb8-a8 Ba6-b7+
38/47 01:49 287,925,092 2,638,000 +M18 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-f5 Kg3-f3 Kf5-f6 Kf3-f4 Kf6-e6 Bf1-c4+ Ke6-d6 Kf4-f5 Kd6-d7 Kf5-e5 Kd7-e7 Bg1-c5+ Ke7-d7 Bc4-b5+ Kd7-c7 Ke5-e6 Kc7-c8 Ke6-d6 Kc8-b7 Kd6-d7 Kb7-b8 Bb5-a6 Kb8-a8 Bc5-d4 Ka8-b8 Kd7-c6 Kb8-a8 Kc6-b6 Ka8-b8 Bd4-e5+ Kb8-a8 Ba6-b7+
39/47 02:22 375,698,922 2,635,000 +M17 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-f5 Kg3-f3 Kf5-e6 Kf3-e4 Ke6-d6 Bf1-c4 Kd6-c6 Ke4-e5 Kc6-d7 Bc4-b5+ Kd7-e7 Bg1-c5+ Ke7-f7 Ke5-f5 Kf7-g7 Bb5-e8 Kg7-h6 Be8-g6 Kh6-g7 Kf5-g5 Kg7-h8 Kg5-f6 Kh8-g8 Bg6-d3 Kg8-h8 Kf6-g6 Kh8-g8 Bd3-c4+ Kg8-h8 Bc5-d4+
40/47 02:49 444,657,564 2,629,000 +M17 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-f5 Kg3-f3 Kf5-e6 Kf3-e4 Ke6-d6 Bf1-c4 Kd6-c6 Ke4-e5 Kc6-d7 Bc4-b5+ Kd7-e7 Bg1-c5+ Ke7-f7 Ke5-f5 Kf7-g7 Bb5-e8 Kg7-h6 Be8-g6 Kh6-g7 Kf5-g5 Kg7-h8 Kg5-f6 Kh8-g8 Bg6-d3 Kg8-h8 Kf6-g6 Kh8-g8 Bd3-c4+ Kg8-h8 Bc5-d4+
41/47 03:26 541,095,950 2,622,000 +M17 Kh1-g2 Kd5-e4 Kg2-g3 Ke4-f5 Kg3-f3 Kf5-e6 Kf3-e4 Ke6-d6 Bf1-c4 Kd6-c6 Ke4-e5 Kc6-d7 Bc4-b5+ Kd7-e7 Bg1-c5+ Ke7-f7 Ke5-f5 Kf7-g7 Bb5-e8 Kg7-h6 Be8-g6 Kh6-g7 Kf5-g5 Kg7-h8 Kg5-f6 Kh8-g8 Bg6-d3 Kg8-h8 Kf6-g6 Kh8-g8 Bd3-c4+ Kg8-h8 Bc5-d4+
houdini3 search and mate scores
Moderator: Ras
-
Uri Blass
- Posts: 10784
- Joined: Thu Mar 09, 2006 12:37 am
- Location: Tel-Aviv Israel
-
syzygy
- Posts: 5674
- Joined: Tue Feb 28, 2012 11:56 pm
Re: houdini3 search and mate scores
It should not, because in this position the answer "draw" would be far more incorrect than the answer "mate (in 85)".Uri Blass wrote:I gave it 1 cpu and 1024 mbytes hash
at depth 23 it found mate in 46
I could expect at bigger depths to see only shorter mates or the same score but houdini shows at depth 24 mate in 48 at depth 25 mate in 65 at depth 26 mate in 75 at depth 27 mate in 80 and at depth 28 mate in 85(note that mate in more than 50 moves means a draw by the 50 move rule so I think that houdini should get draw score for these long mates
This is how transposition tables work. Efficient transposition tables will never work completely correctly with the 50-move rule. But in this case there is not even a problem. If you don't like to see "+M85", then just imagine it says "0.00" there.
-
Uri Blass
- Posts: 10784
- Joined: Thu Mar 09, 2006 12:37 am
- Location: Tel-Aviv Israel
Re: houdini3 search and mate scores
I do not say that houdini should get draw score for the root position but only for long mates.syzygy wrote:It should not, because in this position the answer "draw" would be far more incorrect than the answer "mate (in 85)".Uri Blass wrote:I gave it 1 cpu and 1024 mbytes hash
at depth 23 it found mate in 46
I could expect at bigger depths to see only shorter mates or the same score but houdini shows at depth 24 mate in 48 at depth 25 mate in 65 at depth 26 mate in 75 at depth 27 mate in 80 and at depth 28 mate in 85(note that mate in more than 50 moves means a draw by the 50 move rule so I think that houdini should get draw score for these long mates
This is how transposition tables work. Efficient transposition tables will never work completely correctly with the 50-move rule. But in this case there is not even a problem. If you don't like to see "+M85", then just imagine it says "0.00" there.
The fact that there is a long mate does not mean that there is no shorter mate and the score for the root position is not based only on the long mates but also on shorter lines.
-
syzygy
- Posts: 5674
- Joined: Tue Feb 28, 2012 11:56 pm
Re: houdini3 search and mate scores
Yes, and to that I say: mate (in 85) is far more correct than "draw".Uri Blass wrote:I do not say that houdini should get draw score for the root position but only for long mates.syzygy wrote:It should not, because in this position the answer "draw" would be far more incorrect than the answer "mate (in 85)".Uri Blass wrote:I gave it 1 cpu and 1024 mbytes hash
at depth 23 it found mate in 46
I could expect at bigger depths to see only shorter mates or the same score but houdini shows at depth 24 mate in 48 at depth 25 mate in 65 at depth 26 mate in 75 at depth 27 mate in 80 and at depth 28 mate in 85(note that mate in more than 50 moves means a draw by the 50 move rule so I think that houdini should get draw score for these long mates
This is how transposition tables work. Efficient transposition tables will never work completely correctly with the 50-move rule. But in this case there is not even a problem. If you don't like to see "+M85", then just imagine it says "0.00" there.
In general, finding a mate in 85 does not imply a 50-move rule draw. When an engine finds a mate score in the hash table, say mate in 40, it has no way of knowing whether the path to mate "fits" within the number of moves left. The path to mate will usually have captures or pawn moves.
In this particular case, it is clear that any path to mate will not have pawn moves (since there are no pawns) and will not have captures (since capturing a bishop leads to draw). So yes, an engine could have specialised knowledge for the case KBBvK that will let it evaluate a mate in 85 as a draw. This would be completely stupid, since KBBvK can easily be won within 50 moves.
If you don't want to see a mate in 85 here, then at least ask for a score of +50.00 or so. A draw score is just silly for KBBvK.
Exactly, so a draw score does not make any sense here.The fact that there is a long mate does not mean that there is no shorter mate and the score for the root position is not based only on the long mates but also on shorter lines.
Maybe what you mean is that during the search, if probing the hash table results in mate in 40 and the 50-move counter is already at 10 moves / 20 ply, the search should evaluate the position as a draw. As I have tried to explain, that would be wrong for very many reasons.
The fact of the matter is that the 50-move rule inherently does not play nice with transposition tables. There is no way around this, unless you want to give up most of the benefit of hash tables.
-
michiguel
- Posts: 6401
- Joined: Thu Mar 09, 2006 8:30 pm
- Location: Chicago, Illinois, USA
Re: houdini3 search and mate scores
Mate scores are a very special case of "scores" compared to +9.00 etc. The score per se is just "check mate", meaning, my side wins and the chances to win are 100%. However, since progress is needed, it contains an extra little piece of information, which is how distant that win is. That implies that the true meaning of a mate score is not "Mate in 10", but "Mate in 10 or better". From this perspective, Mate in 85 is not a wrong score.syzygy wrote:Yes, and to that I say: mate (in 85) is far more correct than "draw".Uri Blass wrote:I do not say that houdini should get draw score for the root position but only for long mates.syzygy wrote:It should not, because in this position the answer "draw" would be far more incorrect than the answer "mate (in 85)".Uri Blass wrote:I gave it 1 cpu and 1024 mbytes hash
at depth 23 it found mate in 46
I could expect at bigger depths to see only shorter mates or the same score but houdini shows at depth 24 mate in 48 at depth 25 mate in 65 at depth 26 mate in 75 at depth 27 mate in 80 and at depth 28 mate in 85(note that mate in more than 50 moves means a draw by the 50 move rule so I think that houdini should get draw score for these long mates
This is how transposition tables work. Efficient transposition tables will never work completely correctly with the 50-move rule. But in this case there is not even a problem. If you don't like to see "+M85", then just imagine it says "0.00" there.
In general, finding a mate in 85 does not imply a 50-move rule draw. When an engine finds a mate score in the hash table, say mate in 40, it has no way of knowing whether the path to mate "fits" within the number of moves left. The path to mate will usually have captures or pawn moves.
In this particular case, it is clear that any path to mate will not have pawn moves (since there are no pawns) and will not have captures (since capturing a bishop leads to draw). So yes, an engine could have specialised knowledge for the case KBBvK that will let it evaluate a mate in 85 as a draw. This would be completely stupid, since KBBvK can easily be won within 50 moves.
If you don't want to see a mate in 85 here, then at least ask for a score of +50.00 or so. A draw score is just silly for KBBvK.
Exactly, so a draw score does not make any sense here.The fact that there is a long mate does not mean that there is no shorter mate and the score for the root position is not based only on the long mates but also on shorter lines.
Maybe what you mean is that during the search, if probing the hash table results in mate in 40 and the 50-move counter is already at 10 moves / 20 ply, the search should evaluate the position as a draw. As I have tried to explain, that would be wrong for very many reasons.
The fact of the matter is that the 50-move rule inherently does not play nice with transposition tables. There is no way around this, unless you want to give up most of the benefit of hash tables.
Traditionally, an engine always found shorter and shorter mates, and if it did not do that, some bug was present. With all the current pruning, not anymore. It is possible that a mate is detected, but later it is converted to a longer one (a defensive move was not pruned anymore). In Gaviota I worked a little bit about this but still these weird effects show up. The behavior is much better, but still there is a little hiccup going from mate in 24 to 25.
Code: Select all
30 1: 0.0 +6.06 1.Kg2
130 2: 0.0 +6.06 1.Kg2 Kd6
553 3: 0.0 +6.07 1.Kg2 Kd6 2.Kf3
2013 4: 0.0 +6.06 1.Kg2 Ke4 2.Bc4 Ke5
4594 5: 0.1 +6.08 1.Kg2 Ke4 2.Bc4 Kf4 3.Kf2
13096 6: 0.1 +6.07 1.Kg2 Ke4 2.Kg3 Ke5 3.Bg2 Kf5
28929 7: 0.2 +6.08 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4
37040 8 0.2 +6.08 1.Bd3 Ke5 2.Kg2 Kd6 3.Bh2+ Kd5 4.Kf3
Kd4 5.Be4
63707 8: 0.3 +6.08 1.Bd3 Ke5 2.Kg2 Kd6 3.Bh2+ Kd5 4.Kf3
Kd4 5.Be4
71833 9 0.3 +6.09 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Ke4
121205 9: 0.4 +6.09 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Ke4
135016 10 0.5 +6.09 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Bf1 Ke6
5.Bc4+ Ke5 6.Ke3
200198 10 0.6 +6.10 1.Kg2 Ke5 2.Bh2+ Kd4 3.Kf3 Kd5 4.Ke3
Ke6 5.Ke4 Kf6 6.Bc4
248660 10: 0.7 +6.10 1.Kg2 Ke5 2.Bh2+ Kd4 3.Kf3 Kd5 4.Ke3
Ke6 5.Ke4 Kf6 6.Bc4
297225 11 0.8 +6.10 1.Kg2 Ke5 2.Bh2+ Kd4 3.Kf3 Kd5 4.Ke3
Ke6 5.Ke4 Kf6 6.Bd3 [>]
426039 11: 1.0 +6.10 1.Kg2 Ke5 2.Bh2+ Kd4 3.Kf3 Kd5 4.Ke3
Ke6 5.Ke4 Kf6 6.Bd3 [>]
525286 12 1.2 +6.10 1.Kg2 Ke5 2.Bh2+ Kd4 3.Kf3 Kd5 4.Ke3
Ke6 5.Ke4 Kf6 6.Bd3 [>]
749916 12: 1.5 +6.10 1.Kg2 Ke5 2.Bh2+ Kd4 3.Kf3 Kd5 4.Ke3
Ke6 5.Ke4 Kf6 6.Bd3 [>]
925281 13 1.7 +6.11 1.Kg2 Ke5 2.Bh2+ Kd4 3.Kf3 Kd5 4.Ke3
Ke6 5.Ke4 Kf6 6.Bd3 Ke6 7.Bc4+ Kf6 [<]
1237653 13: 2.1 +6.11 1.Kg2 Ke5 2.Bh2+ Kd4 3.Kf3 Kd5 4.Ke3
Ke6 5.Ke4 Kf6 6.Bd3 Ke6 7.Bc4+ Kf6 [<]
1692629 14 2.6 +6.12 1.Kg2 Ke5 2.Kf3 Kd6 3.Ke4 Ke6 4.Bc4+
Kd6 5.Kf5 Kd7 6.Ke5 Kc6 7.Ke6 Kc7 8.Bb5
2186532 14: 3.1 +6.12 1.Kg2 Ke5 2.Kf3 Kd6 3.Ke4 Ke6 4.Bc4+
Kd6 5.Kf5 Kd7 6.Ke5 Kc6 7.Ke6 Kc7 8.Bb5
2955492 15 3.9 +6.13 1.Kg2 Ke5 2.Kf3 Kd6 3.Ke4 Ke6 4.Bc5 Kf6
5.Bc4 Kg5 6.Be7+ Kg6 7.Ke5 Kg7 8.Kf5
Kh6
3707235 15: 4.7 +6.13 1.Kg2 Ke5 2.Kf3 Kd6 3.Ke4 Ke6 4.Bc5 Kf6
5.Bc4 Kg5 6.Be7+ Kg6 7.Ke5 Kg7 8.Kf5
Kh6
4978931 16 5.9 +6.14 1.Kg2 Ke5 2.Kg3 Kd6 3.Bd3 Kd5 4.Kf4 Kd6
5.Bc4 Kd7 6.Ke5 Kc6 7.Bf1 Kb7 8.Bd3 [>]
5748469 16: 6.6 +6.14 1.Kg2 Ke5 2.Kg3 Kd6 3.Bd3 Kd5 4.Kf4 Kd6
5.Bc4 Kd7 6.Ke5 Kc6 7.Bf1 Kb7 8.Bd3 [>]
8160232 17 8.9 +6.14 1.Kg2 Ke5 2.Kg3 Kd6 3.Bd3 Kd5 4.Kf4 Kd6
5.Bc4 Kd7 6.Ke5 Kc6 7.Bf1 Kb7 8.Bd3 Kc7
[<]
10455918 17: 11.1 +6.14 1.Kg2 Ke5 2.Kg3 Kd6 3.Bd3 Kd5 4.Kf4 Kd6
5.Bc4 Kd7 6.Ke5 Kc6 7.Bf1 Kb7 8.Bd3 Kc7
[<]
12590188 18 13.1 +6.15 1.Kg2 Ke5 2.Kf3 Kd6 3.Bc4 Ke5 4.Bc5 Kf5
5.Bd4 Kg5 6.Ke4 Kg6 7.Ke5 Kg5 8.Be3+
Kg4 9.Be2+ Kg3 10.Ba7 [>]
13942201 18: 14.3 +6.15 1.Kg2 Ke5 2.Kf3 Kd6 3.Bc4 Ke5 4.Bc5 Kf5
5.Bd4 Kg5 6.Ke4 Kg6 7.Ke5 Kg5 8.Be3+
Kg4 9.Be2+ Kg3 10.Ba7 [>]
17439120 19 17.6 +6.16 1.Kg2 Ke5 2.Kf3 Kd6 3.Kf4 Ke6 4.Bc4+
Kd6 5.Kf5 Kc6 6.Ke6 Kc7 7.Bb5 Kb7 8.Kd7
Kb8 9.Ba6 Ka8 10.Kd8 Kb8
19481764 19: 19.5 +6.16 1.Kg2 Ke5 2.Kf3 Kd6 3.Kf4 Ke6 4.Bc4+
Kd6 5.Kf5 Kc6 6.Ke6 Kc7 7.Bb5 Kb7 8.Kd7
Kb8 9.Ba6 Ka8 10.Kd8 Kb8
23690793 20 23.4 +6.19 1.Kg2 Ke5 2.Kg3 Kf5 3.Bd3+ Ke5 4.Ba7
Kd5 5.Kf4 Kd6 6.Be3 Kd5 7.Bg1 Kd6 8.Bc4
Kc6 9.Be3 Kb7 10.Ke5 Kc7 [<]
25155350 20: 24.8 +6.19 1.Kg2 Ke5 2.Kg3 Kf5 3.Bd3+ Ke5 4.Ba7
Kd5 5.Kf4 Kd6 6.Be3 Kd5 7.Bg1 Kd6 8.Bc4
Kc6 9.Be3 Kb7 10.Ke5 Kc7 [<]
43508372 21 41.8 +6.19 1.Kg2 Ke5 2.Kg3 Kf5 3.Bd3+ Ke5 4.Ba7
Kd5 5.Kf4 Kd6 6.Be3 Kd5 7.Kf5 Kd6 8.Bc4
Ke7 9.Bc5+ Kd7 10.Bb5+ Kc7 11.Ke6 Kb7
12.Bd4 [>]
47294606 21 45.3 :-) 1.Bg2+
52420172 21 50.1 +Mat_31 1.Bg2+ Kc4 2.Kh2 Kb5 3.Kg3 Kc4 4.Kf4
Kc3 5.Be4 Kb4 6.Ke5 Kb3 7.Kd4 Ka4 8.Be3
Ka3 9.Kd5 [>]
54275576 21 51.9 +Mat_30 1.Kh2 Ke5 2.Kg3 Kf5 3.Bd3+ Ke5 4.Kf3
Kd5 5.Kf4 Ke6 6.Be4 Kd6 7.Kf5 Ke7 8.Ke5
Kd7 9.Bd4 Ke7 [<]
55851576 21: 53.3 +Mat_30 1.Kh2 Ke5 2.Kg3 Kf5 3.Bd3+ Ke5 4.Kf3
Kd5 5.Kf4 Ke6 6.Be4 Kd6 7.Kf5 Ke7 8.Ke5
Kd7 9.Bd4 Ke7 [<]
57187824 22 54.6 +Mat_25 1.Kh2 Ke5 2.Kg3 Kf5 3.Kf3 Ke5 4.Bc4 Kd6
5.Ke4 Kc6 6.Ke5 Kd7 7.Bd3 Ke7 8.Bc5+
Kd7 9.Bb5+ Kc7 10.Ke6 Kb7 [<]
58842275 22: 56.1 +Mat_25 1.Kh2 Ke5 2.Kg3 Kf5 3.Kf3 Ke5 4.Bc4 Kd6
5.Ke4 Kc6 6.Ke5 Kd7 7.Bd3 Ke7 8.Bc5+
Kd7 9.Bb5+ Kc7 10.Ke6 Kb7 [<]
60509241 23 57.7 +Mat_24 1.Kh2 Ke5 2.Kg3 Kf5 3.Bg2 Ke5 4.Kg4 Ke6
5.Bc5 Kd7 6.Kf5 Kc7 7.Bf3 Kd7 8.Ke5 Kc7
9.Ke6 Kb8 [<]
62435237 23: 59.5 +Mat_24 1.Kh2 Ke5 2.Kg3 Kf5 3.Bg2 Ke5 4.Kg4 Ke6
5.Bc5 Kd7 6.Kf5 Kc7 7.Bf3 Kd7 8.Ke5 Kc7
9.Ke6 Kb8 [<]
64670204 24 61.6 +Mat_25 1.Kh2 Ke5 2.Kg3 Kf5 3.Bg2 Ke5 4.Kg4 Ke6
5.Bc5 Kf6 6.Kf4 Ke6 7.Bc6 Kf6 8.Bd7 Kg6
9.Bb5 Kh6 10.Bd3 Kh5 [<]
67549980 24 64.3 +Mat_24 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kc6 4.Ke4 Kd6
5.Bd4 Kc6 6.Ke5 Kd7 7.Bc4 Kc6 8.Ke6 Kc7
9.Bd5 Kd8 10.Bb7 Kc7 [<]
68083942 24: 64.8 +Mat_24 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kc6 4.Ke4 Kd6
5.Bd4 Kc6 6.Ke5 Kd7 7.Bc4 Kc6 8.Ke6 Kc7
9.Bd5 Kd8 10.Bb7 Kc7 [<]
69012443 25 65.7 +Mat_23 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kc6 4.Ke4 Kd6
5.Bd4 Kc6 6.Ke5 Kd7 7.Bb5+ Kc7 8.Ke6
Kb7 9.Kd6 Kb8 10.Kc6 Kc8 11.Bd3 [>]
72043417 25: 68.5 +Mat_23 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kc6 4.Ke4 Kd6
5.Bd4 Kc6 6.Ke5 Kd7 7.Bb5+ Kc7 8.Ke6
Kb7 9.Kd6 Kb8 10.Kc6 Kc8 11.Bd3 [>]
75724350 26 72.0 +Mat_22 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kc6 4.Ke4 Kd6
5.Bd4 Kc6 6.Ke5 Kd7 7.Be4 Kc7 8.Ke6 Kc8
9.Be5 Kd8 10.Bd3 Kc8 11.Ba6+ Kd8 12.Bd6
[>]
80034591 26: 76.1 +Mat_22 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kc6 4.Ke4 Kd6
5.Bd4 Kc6 6.Ke5 Kd7 7.Be4 Kc7 8.Ke6 Kc8
9.Be5 Kd8 10.Bd3 Kc8 11.Ba6+ Kd8 12.Bd6
[>]
81656748 27 77.6 +Mat_20 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kc6 4.Ke4 Kd6
5.Bd4 Kc6 6.Ke5 Kd7 7.Be4 Ke7 8.Bf5 Kf7
9.Kd6 Ke8 10.Bg6+ Kd8 11.Bc5 Kc8 12.Kc6
Kd8 [<]
86692106 27: 82.4 +Mat_20 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kc6 4.Ke4 Kd6
5.Bd4 Kc6 6.Ke5 Kd7 7.Be4 Ke7 8.Bf5 Kf7
9.Kd6 Ke8 10.Bg6+ Kd8 11.Bc5 Kc8 12.Kc6
Kd8 [<]
100163079 28 95.1 +Mat_20 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kc6 4.Ke4 Kd6
5.Bc4 Kc6 6.Ke5 Kd7 7.Bb5+ Ke7 8.Bc5+
Kf7 9.Kf5 Kg7 10.Be8 Kh7 11.Kf6 Kh6
12.Bd4 Kh7 13.Be3 [>]
106359465 28: 101.0 +Mat_20 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kc6 4.Ke4 Kd6
5.Bc4 Kc6 6.Ke5 Kd7 7.Bb5+ Ke7 8.Bc5+
Kf7 9.Kf5 Kg7 10.Be8 Kh7 11.Kf6 Kh6
12.Bd4 Kh7 13.Be3 [>]
126711350 29 120.2 +Mat_19 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Be4+
Ke5 5.Bc5 Kf6 6.Kf4 Ke6 7.Bf3 Kd7 8.Ke5
Kd8 9.Kd6 Ke8 10.Bh5+ Kf8 11.Ke6+ Kg7
12.Be3 Kf8 13.Bh6+ Kg8 14.Bg6 Kh8
15.Kf6 Kg8 [<]
135650296 29: 128.5 +Mat_19 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Be4+
Ke5 5.Bc5 Kf6 6.Kf4 Ke6 7.Bf3 Kd7 8.Ke5
Kd8 9.Kd6 Ke8 10.Bh5+ Kf8 11.Ke6+ Kg7
12.Be3 Kf8 13.Bh6+ Kg8 14.Bg6 Kh8
15.Kf6 Kg8 [<]
136893219 30 129.7 +Mat_19 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Be4+
Ke5 5.Bc5 Kf6 6.Kf4 Ke6 7.Bf3 Kd7 8.Ke5
Kd8 9.Kd6 Ke8 10.Bh5+ Kf8 11.Ke6+ Kg7
12.Bf7 Kh6 13.Kf6 Kh7 14.Be3 Kh8 [<]
142382101 30: 134.8 +Mat_19 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Be4+
Ke5 5.Bc5 Kf6 6.Kf4 Ke6 7.Bf3 Kd7 8.Ke5
Kd8 9.Kd6 Ke8 10.Bh5+ Kf8 11.Ke6+ Kg7
12.Bf7 Kh6 13.Kf6 Kh7 14.Be3 Kh8 [<]
144689826 31 137.0 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
149330487 31: 141.3 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
150763422 32 142.7 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
156629269 32: 148.2 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
178688284 33 168.8 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
187760101 33: 177.2 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
189231589 34 178.6 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
195859820 34: 184.7 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
197361375 35 186.1 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
204964850 35: 193.1 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
206476971 36 194.4 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
214600204 36: 201.8 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
216149217 37 203.2 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
223931424 37: 210.3 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
225750218 38 211.9 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
232985165 38: 218.4 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
235505688 39 220.6 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
244464068 39: 228.6 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
247629605 40 231.3 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
255122771 40: 238.1 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
256632541 41 239.4 +Mat_17 1.Bd3 Ke5 2.Kg2 Kd6 3.Kf3 Kd5 4.Kf4 Ke6
5.Bc4+ Kd6 6.Be3 Kc6 7.Ke5 Kd7 8.Bb5+
Ke7 9.Bg5+ Kf7 10.Kf5 Kg7 11.Bc4 Kf8
12.Kg6 Ke8 13.Bb5+ Kf8 14.Bc6 Kg8
15.Bh6 Kh8 16.Bg7+ Kg8 17.Bd5#
-
syzygy
- Posts: 5674
- Joined: Tue Feb 28, 2012 11:56 pm
Re: houdini3 search and mate scores
I agree.michiguel wrote:Mate scores are a very special case of "scores" compared to +9.00 etc. The score per se is just "check mate", meaning, my side wins and the chances to win are 100%. However, since progress is needed, it contains an extra little piece of information, which is how distant that win is. That implies that the true meaning of a mate score is not "Mate in 10", but "Mate in 10 or better". From this perspective, Mate in 85 is not a wrong score.
Here I am a bit doubtful. I suppose most if not all engines do not prune moves that could defend against a mate score (or if they do, their authors would consider it a bug). The next ply finding only longer mates is probably the result of transposition table entries being overwritten. Another possibility is that for whatever reason the next ply reduces some moves more than the present ply (or prunes more moves from the winning side).Traditionally, an engine always found shorter and shorter mates, and if it did not do that, some bug was present. With all the current pruning, not anymore. It is possible that a mate is detected, but later it is converted to a longer one (a defensive move was not pruned anymore).
My engine happens to have no hiccups on this position. First mate in 20, then 19, 18, 17. But it is certainly not hiccup free in general.In Gaviota I worked a little bit about this but still these weird effects show up. The behavior is much better, but still there is a little hiccup going from mate in 24 to 25.
-
rabbits
- Posts: 276
- Joined: Wed Mar 08, 2006 10:14 pm
- Location: Sydney, Australia
Re: houdini3 search and mate scores
Chessmaster 10 announces mate in 17 in five seconds.
Al
Al
-
Mike S.
- Posts: 1480
- Joined: Thu Mar 09, 2006 5:33 am
Re: houdini3 search and mate scores
From Houdini 1.5a(*) without Gaviotas, I got this:
Mate in 18 [+] 1. Bd3 Ke5 2. Kg2 Kd5 3. Kf3 Kd6 4. Ke4 Ke6 5. Bc4+ Kd6 6. Be3 Kd7 7. Ke5 Ke7 8. Bc5+ Kd7 9. Bb5+ Kc7 10. Ke6 Kc8 11. Kd6 Kd8 12. Bb6+ Kc8 13. Ba6+ Kb8 14. Bg1 Ka8 15. Kc6 Kb8 16. Kb6 Ka8 17. Bb7+ Kb8 18. Bh2# (depth 29, 0:01:26)
[+] Tablebase: 1.Bd3 - White wins in 17 moves
*) with internal 2.0c-settings, but that probably doesn't matter here
Mate in 18 [+] 1. Bd3 Ke5 2. Kg2 Kd5 3. Kf3 Kd6 4. Ke4 Ke6 5. Bc4+ Kd6 6. Be3 Kd7 7. Ke5 Ke7 8. Bc5+ Kd7 9. Bb5+ Kc7 10. Ke6 Kc8 11. Kd6 Kd8 12. Bb6+ Kc8 13. Ba6+ Kb8 14. Bg1 Ka8 15. Kc6 Kb8 16. Kb6 Ka8 17. Bb7+ Kb8 18. Bh2# (depth 29, 0:01:26)
[+] Tablebase: 1.Bd3 - White wins in 17 moves
*) with internal 2.0c-settings, but that probably doesn't matter here
Regards, Mike