I've encountered a bug in the Pikafish Xiangqi engine where the engine stops calculating moves after reaching depth 15-16 for a specfic position. This issue arises when using a full FEN string command but does not occur with a shortened version of the FEN string.
Expected behavior
The engine should consistently calculate up to the specified depth (in this case, depth 18) regardless of whether additional parameters (like halfmove clock and fullmove number) are included in the FEN string.
Steps to reproduce
Input the command with a full FEN string:
position fen 1rbP1a3/2N1k4/4b1n2/p3p3p/7P1/2C5P/P3P4/B1N1C4/4Ar3/3RKAB1c b - - 2 26
go depth 18
Observe that the engine stops calculating between depths 15-16.
Input the command with a shortened FEN string (without the additional parameters after the move side):
position fen 1rbP1a3/2N1k4/4b1n2/p3p3p/7P1/2C5P/P3P4/B1N1C4/4Ar3/3RKAB1c b
go depth 18
Observe that the engine calculates as expected up to depth 18.
Anything else?
It's unclear why removing the " - - 2 26" part from the FEN string resolves the issue. This inconsistency in behavior depending on the FEN string format is puzzling and seems to be a bug.
Describe the issue
I've encountered a bug in the Pikafish Xiangqi engine where the engine stops calculating moves after reaching depth 15-16 for a specfic position. This issue arises when using a full FEN string command but does not occur with a shortened version of the FEN string.
Expected behavior
The engine should consistently calculate up to the specified depth (in this case, depth 18) regardless of whether additional parameters (like halfmove clock and fullmove number) are included in the FEN string.
Steps to reproduce
Input the command with a full FEN string: position fen 1rbP1a3/2N1k4/4b1n2/p3p3p/7P1/2C5P/P3P4/B1N1C4/4Ar3/3RKAB1c b - - 2 26 go depth 18 Observe that the engine stops calculating between depths 15-16.
Input the command with a shortened FEN string (without the additional parameters after the move side): position fen 1rbP1a3/2N1k4/4b1n2/p3p3p/7P1/2C5P/P3P4/B1N1C4/4Ar3/3RKAB1c b go depth 18 Observe that the engine calculates as expected up to depth 18.
Anything else?
It's unclear why removing the " - - 2 26" part from the FEN string resolves the issue. This inconsistency in behavior depending on the FEN string format is puzzling and seems to be a bug.
Operating system
Windows
Pikafish version
Pikafish 2023-03-05