Mate in 3: 7 minutes for SF 17

[Jouni]

[d]8/8/5k2/p6p/P1r2n1K/8/8/7R b - - 0 1
This was posted in discord.
1...Rc3 2.Rh2 Kg6 3.Rh1 Ng2#
-+ (-#3) Depth: 48/62 00:07:44 1426mN
There are other programs to struggle like Obsidian.

[cc2150dx]

or:

[d]8/8/5k2/p6p/P1r2n1K/8/8/7R b - - 0 1

Lc0:

3/8 00:00 166 1k -M3 1. ... Rc3 2.Rh2 Kg6 3.Rg2+ Nxg2+

[peter]

[d]8/8/5k2/p6p/P1r2n1K/8/8/7R b - - 0 1

3.5GHz- CPU, hyperthreading on:

Code: Select all

Stockfish 17 by the Stockfish developers (see AUTHORS file)
position fen 8/8/5k2/p6p/P1r2n1K/8/8/7R b - - 0 1
setoption name hash value 1024
go movetime 100000
info string Available processors: 0-31
info string Using 1 thread
info string NNUE evaluation using nn-1111cefa1111.nnue (133MiB, (22528, 3072, 15, 32, 1))
info string NNUE evaluation using nn-37f18f62d772.nnue (6MiB, (22528, 128, 15, 32, 1))
...
info depth 42 seldepth 83 multipv 1 score mate 3 nodes 54363201 nps 3425316 hashfull 108 tbhits 0 time 15871 pv c4c3 h1h2 f6g6 h2h1 f4g2

[Brunetti]

Code: Select all

info depth 42 seldepth 83 multipv 1 score mate 3 nodes 54363201 nps 3425316 hashfull 108 tbhits 0 time 15871 pv c4c3 h1h2 f6g6 h2h1 f4g2

Quite random behavior. With any other hash value between 16 and 8192, it takes more time. And increasing the threads also increases the time!

Alex

[peter]

That hash matters so much even single threaded is weird indeed, that multithread differs randomly isn't as big a surprise.
One more trial:

Code: Select all

Stockfish 17 by the Stockfish developers (see AUTHORS file)
position fen 8/8/5k2/p6p/P1r2n1K/8/8/7R b - - 0 1
setoption name hash value 1024
setoption name threads value 2
info string Using 2 threads
go movetime 100000
info string Available processors: 0-31
info string Using 2 threads
info string NNUE evaluation using nn-1111cefa1111.nnue (133MiB, (22528, 3072, 15, 32, 1))
info string NNUE evaluation using nn-37f18f62d772.nnue (6MiB, (22528, 128, 15, 32, 1))
...
info depth 36 seldepth 58 multipv 1 score mate 3 nodes 194618758 nps 7232747 hashfull 271 tbhits 0 time 26908 pv c4c3 h1h2 f6g6 h2h1 f4g2

[Brunetti]

That hash matters so much even single threaded is weird indeed, that multithread differs randomly isn't as big a surprise.

Yes, but with multipv everything changes

Code: Select all

position fen 8/8/5k2/p6p/P1r2n1K/8/8/7R b - - 0 1
setoption name hash value 8192
setoption name Threads value 16
setoption name multipv value 10
go movetime 1000
...
info depth 16 seldepth 6 multipv 1 score mate 3 nodes 8457346 nps 21356934 hashfull 5 tbhits 0 time 396 pv c4c3 h1h2 f6g6 h2h1 f4g2

Alex

[chessica]

[d]8/8/5k2/p6p/P1r2n1K/8/8/7R b - - 0 1

Stockfish_11:

00:06,844 3.103k 453k +M3 Tc4-c3 Th1-h2 Kf6-g6 Th2-b2 Tc3-h3+


Stockfish_14.1:

30/33+ 00:03,088 483k 156k +M3 Tc4-c3


Stockfish-15.1:

40/41 00:08,994 3.610k 401k +M3 Tc4-c3 Th1-h2 Kf6-g6 Th2-h1 Sf4-g2+


Stockfish-16.1:

32/37 02:16,203 47.369k 348k +M3 Tc4-c3 Th1-h2 Kf6-g6 Th2-h1 Sf4-g2+


Stockfish-17:

40/62 02:56,769 109.731k 621k +M3 Tc4-c3 Th1-h2 Kf6-g6 Th2-h1 Sf4-g2+

[hgm]

Interesting. Fairy-Max finds the mate after 39 sec, at 21 ply. Normally a mate in 3 would require 7 ply (seen when the King gets captured). And it does not use excessive reductions. So it should see it quickly, or not at all. In this case I had expected the 'not at all', because the main line contains a zugzwang, and Fairy-Max does not do null-move verification. So I would expect black to happily defend against the mate by passing his turn to leave the Rook on h2, to protect the mate square g2.

Apparently this defense can be punished only after some 10 moves, perhaps by promoting the a-Pawn. The higher root score would then make the null moves after 1... Rc3 and 2... Kg6 fail low, so that white is forced to allow the mate, and the mentioned moves are recognized as the best.

Perhaps the fact that it takes quite some moves to punish the null moves is also the explanation why Stockfish has trouble finding the mate.

21 #3 79.9M 0:38.90 c4c3 h1h2 f6g6 h2h1 f4g2
21 +4.72 74.3M 0:36.20 f6f5 h4g3 c4a4 h1b1 a4a3 g3f2 h5h4 b1b7 f4d3 f2g1 h4h3 b7f7 f5e4 f7h7 d3f4 h7e7
20 +4.70 63.1M 0:30.63 f6f5 h4g3 c4a4 h1b1 a4a3 g3h2 a5a4
20 +4.65 51.6M 0:25.13 c4a4 h1b1 a4b4 b1d1 a5a4 d1d6 f6f5 d6d7 a4a3 d7f7 f5e5 h4g5 a3a2 f7f5 e5d6 f5a5 b4b5 a5b5 a2a1 g5f4
19 +4.64 30.1M 0:14.60 c4a4 h4g3 f6f5 g3f3 a4a3 f3f2 a5a4 h1d1 a3c3 d1d4 a4a3 d4a4 f5e5 a4a5 e5e4 a5a4 e4f5 a4a5 f5e6
18 +4.63 8.82M 0:04.24 c4a4 h1d1 f4g6 h4g3 a4a3 g3g2 h5h4 d1f1 f6g5 f1g1 a3g3 g2h2 g3g1 h2g1 h4h3 g1h2 g5g4 h2g1
17 +4.63 5.57M 0:02.68 c4a4 h1d1 f4g6 h4g3 a4a3 g3g2 h5h4 d1f1 f6g5 f1g1 h4h3 g2h2 g5f5 g1g3 a3g3 h2g3 g6f4
16 +4.60 3.54M 0:01.71 c4a4 h1d1 f4g6 h4g3 h5h4 g3h3 a4a3 h3g4 a3g3 g4h5 g3g5 h5h6 h4h3 d1f1 g5f5 f1g1
15 +4.60 2.43M 0:01.17 c4a4 h1d1 f4g6 h4g3 h5h4 g3h3 a4a3 h3g4 a3g3 g4h5 g3g5 h5h6 h4h3 d1f1 g5f5
14 +4.57 1.45M 0:00.71 c4a4 h1b1 a4b4 b1d1 a5a4 d1d6 f6f5 d6a6 f4g6 h4g3 b4b3 g3f2 a4a3 f2e2
13 +4.07 635608 0:00.32 c4a4 h1b1 f4g6 h4h3 g6e5 h3g2 h5h4 b1h1 a4f4 h1a1 a5a4 a1a2 f6f5
12 +4.07 403612 0:00.21 c4a4 h1b1 f4g6 h4h3 g6e5 h3g2 h5h4 b1h1 a4f4 h1a1 a5a4 a1b1
11 +4.07 257270 0:00.14 c4a4 h1b1 f4e6 h4h3 a4g4 b1b6 a5a4 b6a6 f6f5 a6a5 f5g6
10 +4.07 68184 0:00.06 c4a4 h4g3 f6e5 g3f3 a4a3 f3f2 a5a4 h1f1 h5h4 f1g1
9 +4.07 31742 0:00.03 c4a4 h4g3 f6e5 g3f3 a4a3 f3f2 a5a4 h1g1 h5h4
8 +4.07 23056 0:00.01 c4a4 h4g3 f6e5 g3f3 a4a3 f3f2 a5a4 h1g1
7 +4.07 11051 0:00.01 c4a4 h4g3 f6e5 g3f3 a4a3 f3f2 a5a4
6 +4.07 9955 0:00.01 c4a4 h4g3 f6e5 g3f3 a4a3 f3f2
6 +4.06 8904 0:00.01 c4c3 h1h2 c3a3 h2h1 f6f5 h1g1 a3a4
6 +4.01 5051 0:00.00 f4e2 h4h3 c4a4 h3g2 e2f4 g2f3
5 +4.06 2584 0:00.00 f4e2 h4h3 c4a4 h3g2
5 +4.04 2358 0:00.00 c4a4 h4g3 f6e5 g3f3 a4b4
4 +4.04 526 0:00.00 c4a4 h4g3 f6e5 g3f3
3 +4.07 308 0:00.00 c4a4 h4g3 f6e5
2 +4.04 17 0:00.00 c4a4 h4g3
1 +4.07 8 0:00.00 c4a4
0 #

[JohnW]

It took Stockfish 17 about 30 seconds on my 10 year old pc

[chessica]

[d]8/8/5k2/p6p/P1r2n1K/8/8/7R b - - 0 1

Weiss 2.1-dev.16e859e JA:

10/29+ 00:03,262 5.495k 1.684k +M3 Tc4-c3
10/29 00:03,262 5.495k 1.684k +M3 Tc4-c3 Th1-h2 Kf6-g6 Th2-b2 Tc3-h3+