Search explosion

[PikaCat-OuO]

Describe the issue

7k/7P/8/8/4B3/3B4/2B5/KB6 w - - 0 1
Search explosion on very low depth, and stuck with no progress.
Possibly has something to do with b5f1108
Position found by @xiangqishare
The explosion is easier to trigger with more search threads.

Expected behavior

Not overdo extension maybe?

Steps to reproduce

Stockfish dev-20250629-15555e8f by the Stockfish developers (see AUTHORS file)
setoption name Threads value 32
info string Using 32 threads
setoption name Hash value 8192
position fen 7k/7P/8/8/4B3/3B4/2B5/KB6 w - - 0 1
go
info string Available processors: 0-31
info string Using 32 threads
info string NNUE evaluation using nn-1c0000000000.nnue (133MiB, (22528, 3072, 15, 32, 1))
info string NNUE evaluation using nn-37f18f62d772.nnue (6MiB, (22528, 128, 15, 32, 1))
info depth 1 seldepth 2 multipv 1 score cp 555 nodes 1023 nps 511500 hashfull 0 tbhits 0 time 2 pv a1a2
info depth 2 seldepth 5 multipv 1 score cp 612 nodes 79443 nps 9930375 hashfull 0 tbhits 0 time 8 pv c2a4 h8g7
info depth 3 seldepth 5 multipv 1 score cp 665 nodes 82908 nps 10363500 hashfull 0 tbhits 0 time 8 pv a1a2
info depth 4 seldepth 5 multipv 1 score cp 706 nodes 87259 nps 9695444 hashfull 0 tbhits 0 time 9 pv a1a2 h8g7 d3b5
info depth 5 seldepth 6 multipv 1 score cp 675 nodes 90965 nps 10107222 hashfull 0 tbhits 0 time 9 pv a1a2 h8g7 d3b5 g7h8 a2b3
info depth 6 seldepth 9 multipv 1 score cp 678 nodes 94762 nps 10529111 hashfull 0 tbhits 0 time 9 pv a1a2 h8g7 a2b2
info depth 7 seldepth 8 multipv 1 score cp 669 nodes 99587 nps 9958700 hashfull 0 tbhits 0 time 10 pv a1b2 h8g7 h7h8b g7h8 d3b5
info depth 8 seldepth 13 multipv 1 score cp 684 nodes 109600 nps 9963636 hashfull 0 tbhits 0 time 11 pv a1b2 h8g7 b2c3 g7h8 c3b3 h8g7 b3c4 g7h8 c4b5 h8g7
(search hangs here)

Anything else?

No response

Operating system

All

Stockfish version

15555e8

[robertnurnberg]

I can confirm this. Might be of interest to @xu-shawn, not least as another position to add to our collection of positions that trigger such explosions.

[dav1312]

#5023

[Matthies]

German user reports search explosion (well, he doesn't call it search explosion, just "Stockfish unbrauchbar und ein UpDate ist dringend erforderlich") with new network architecture. Probably not really related to new network but worth to mention here:

$ ./stockfish
Stockfish dev-20251112-8e5392d7 by the Stockfish developers (see AUTHORS file)
setoption name Hash value 2048
position fen 8/1pp2k2/p2p3p/3n1KP1/5P1P/8/1P1B4/8 w - - 0 1
go
info string Available processors: 0-15
info string Using 1 thread
info string NNUE evaluation using nn-49c1193b131c.nnue (125MiB, (102384, 1024, 15, 32, 1))
info string NNUE evaluation using nn-37f18f62d772.nnue (6MiB, (22528, 128, 15, 32, 1))
info string Network replica 1: Shared memory.
info depth 1 seldepth 4 multipv 1 score cp 200 nodes 17 nps 17000 hashfull 0 tbhits 0 time 1 pv g5g6 f7e7
…
info depth 38 seldepth 76 multipv 1 score cp 43 lowerbound nodes 31698274 nps 1213749 hashfull 54 tbhits 0 time 26116 pv h4h5
info depth 37 currmove h4h5 currmovenumber 1
info depth 38 seldepth 76 multipv 1 score cp 48 lowerbound nodes 31962598 nps 1215631 hashfull 54 tbhits 0 time 26293 pv h4h5
info depth 36 currmove h4h5 currmovenumber 1
info depth 38 seldepth 126 multipv 1 score cp 54 lowerbound nodes 435940215 nps 2801780 hashfull 59 tbhits 0 time 155594 pv h4h5
info depth 35 currmove h4h5 currmovenumber 1
info depth 38 seldepth 142 multipv 1 score cp 61 lowerbound nodes 1897823722 nps 3023133 hashfull 60 tbhits 0 time 627767 pv h4h5
info depth 34 currmove h4h5 currmovenumber 1
info depth 38 seldepth 142 multipv 1 score cp 71 lowerbound nodes 2295553559 nps 2947646 hashfull 61 tbhits 0 time 778775 pv h4h5
info depth 33 currmove h4h5 currmovenumber 1
info depth 33 currmove g5h6 currmovenumber 2
info depth 33 currmove g5g6 currmovenumber 3
info depth 33 currmove d2c3 currmovenumber 4
info depth 33 currmove f5e4 currmovenumber 5
info depth 33 currmove d2e1 currmovenumber 6
info depth 33 currmove f5g4 currmovenumber 7
info depth 33 currmove d2a5 currmovenumber 8
info depth 33 currmove d2e3 currmovenumber 9
info depth 33 currmove b2b3 currmovenumber 10
info depth 33 currmove d2c1 currmovenumber 11
info depth 33 currmove b2b4 currmovenumber 12
info depth 33 currmove d2b4 currmovenumber 13
info depth 38 seldepth 142 multipv 1 score cp 62 nodes 23075245840 nps 1424077 hashfull 64 tbhits 0 time 16203640 pv h4h5 d5e7 f5e4 c7c6 d2e1 e7d5 g5h6 b7b5 e1a5 d5f6 e4f5 c6c5 a5c7 d6d5 c7d6 c5c4 d6e5 f6h7 e5c3 a6a5 c3a5 d5d4 f5e5 d4d3 a5c3 h7f6 e5d4 f7g8 c3d2 g8f7 f4f5 f7g8 d2e1 g8f7 e1c3 f7g8 c3b4

[dav1312]

German user reports search explosion (well, he doesn't call it search explosion, just "Stockfish unbrauchbar und ein UpDate ist dringend erforderlich") with new network architecture. Probably not really related to new network but worth to mention here:

They've been posting that position for a few days because the new net is "blind" to it. I don't think this is related to the explosions.

[Matthies]

They've been posting that position for a few days because the new net is "blind" to it. I don't think this is related to the explosions.

seldepth 142
Still running here on depth 33. Will update output in my post above when next line is printed... going to sleep now... next output after 4,5h.