stockfish.info returning mate in 1 in a checkmate position

[guidopetri]

Example code:

sf = stockfish.Stockfish(stockfish_executable)
sf.set_fen_position('8/8/8/6pp/8/4k1PP/8/r3K3 w - - 12 53')
sf.get_best_move()
sf.info  # blah blah mate 1

The position in the FEN is checkmate, but info string returns a mate in 1:

'info depth 10 seldepth 2 multipv 1 score mate 1 nodes 204 nps 204000 tbhits 0 time 1 pv a2a1'

The FEN position is the last position in this game.

[guidopetri]

This doesn't seem to be the case for the library version 3.0.0. In that version, sf.info returns ''.

[guidopetri]

Sorry, never mind - bug is still present in 3.0.0. I must have never implemented the functionality correctly...

[zhelyabuzhsky]

Hi, @charlesoblack

I wrote test:

def test_set_fen_position_mate(self, stockfish):
    stockfish.set_fen_position('8/8/8/6pp/8/4k1PP/8/r3K3 w - - 12 53')
    assert stockfish.get_best_move() is None
    assert stockfish.info == ''

I think it's correct behaviour. Do you agree?

[guidopetri]

Yes, I think that test displays the intended behavior.

[zhelyabuzhsky]

So ... no bug? Can I close this issue?

[guidopetri]

While the test itself passes for me, the current behavior still shows up when running through a game with a python.chess Visitor:

class TestVisitor(chess.pgn.BaseVisitor):
    
    def __init__(self, gm):
        self.gm = gm
        self.gm.evals = []
        self.move_count = 1

    def visit_board(self, board):
        self.move_count += 1
        sf.set_fen_position(board.fen())
        best_move = sf.get_best_move()
        print(best_move, sf.info)
        info_string = sf.info
        rating_match = re.search(r'score (cp|mate) (.+?)(?: |$)', info_string)
        if rating_match.group(1) == 'mate':
            original_rating = int(rating_match.group(2))
            if original_rating:
                rating = 9999 * original_rating / abs(original_rating)
            elif self.gm.headers['Result'] == '1-0':
                rating = 9999
            else:
                rating = -9999
        else:
            rating = int(rating_match.group(2))
        if board.turn == chess.BLACK:
            rating *= -1
        self.gm.evals.append(rating)

output ends with:

a2a1 info depth 10 seldepth 2 multipv 1 score mate 1 nodes 204 nps 204000 tbhits 0 time 1 pv a2a1
None info depth 10 seldepth 2 multipv 1 score mate 1 nodes 204 nps 204000 tbhits 0 time 1 pv a2a1

Here is the PGN I am using for the above:

[Event "Rated Blitz game"]
[Site "https://lichess.org/3N2cqCjY"]
[Date "2020.02.14"]
[Round "-"]
[White "Grahtbo"]
[Black "Gorbunok"]
[Result "0-1"]
[BlackElo "1734"]
[BlackRatingDiff "+6"]
[ECO "A13"]
[Opening "English Opening: Agincourt Defense"]
[Termination "Normal"]
[TimeControl "180+0"]
[UTCDate "2020.02.14"]
[UTCTime "22:08:23"]
[Variant "Standard"]
[WhiteElo "1748"]
[WhiteRatingDiff "-6"]

1. c4 { [%clk 0:03:00] } 1... e6 { [%clk 0:03:00] } 2. b3 { [%clk 0:02:59] } 2... c6 { [%clk 0:02:59] } 3. a3 { [%clk 0:02:58] } 3... a5 { [%clk 0:02:58] } 4. Bb2 { [%clk 0:02:57] } 4... Nf6 { [%clk 0:02:55] } 5. e3 { [%clk 0:02:56] } 5... Be7 { [%clk 0:02:54] } 6. d3 { [%clk 0:02:55] } 6... O-O { [%clk 0:02:53] } 7. Nf3 { [%clk 0:02:54] } 7... h6 { [%clk 0:02:51] } 8. Be2 { [%clk 0:02:54] } 8... b6 { [%clk 0:02:50] } 9. O-O { [%clk 0:02:53] } 9... Ba6 { [%clk 0:02:48] } 10. Nbd2 { [%clk 0:02:52] } 10... d5 { [%clk 0:02:46] } 11. cxd5 { [%clk 0:02:51] } 11... cxd5 { [%clk 0:02:45] } 12. Qc2 { [%clk 0:02:50] } 12... Nbd7 { [%clk 0:02:43] } 13. Rac1 { [%clk 0:02:49] } 13... Rc8 { [%clk 0:02:41] } 14. Qb1 { [%clk 0:02:48] } 14... Rxc1 { [%clk 0:02:39] } 15. Qxc1 { [%clk 0:02:45] } 15... e5 { [%clk 0:02:26] } 16. Nxe5 { [%clk 0:02:41] } 16... Nxe5 { [%clk 0:02:25] } 17. Bxe5 { [%clk 0:02:40] } 17... Bd6 { [%clk 0:02:21] } 18. Bb2 { [%clk 0:02:38] } 18... Re8 { [%clk 0:02:18] } 19. Qa1 { [%clk 0:02:34] } 19... Bb7 { [%clk 0:02:07] } 20. Nf3 { [%clk 0:02:31] } 20... d4 { [%clk 0:02:00] } 21. Bxd4 { [%clk 0:02:25] } 21... Bxf3 { [%clk 0:01:57] } 22. Bxf3 { [%clk 0:02:24] } 22... Qe7 { [%clk 0:01:40] } 23. e4 { [%clk 0:02:22] } 23... Rc8 { [%clk 0:01:33] } 24. Bxb6 { [%clk 0:02:21] } 24... Be5 { [%clk 0:01:29] } 25. Bd4 { [%clk 0:02:17] } 25... Bxd4 { [%clk 0:01:28] } 26. Qxd4 { [%clk 0:02:15] } 26... Rc2 { [%clk 0:01:23] } 27. b4 { [%clk 0:02:07] } 27... axb4 { [%clk 0:01:19] } 28. Qxb4 { [%clk 0:02:06] } 28... Qxb4 { [%clk 0:01:15] } 29. axb4 { [%clk 0:02:05] } 29... Rb2 { [%clk 0:01:14] } 30. e5 { [%clk 0:02:03] } 30... Nd7 { [%clk 0:01:12] } 31. Bc6 { [%clk 0:02:01] } 31... Nxe5 { [%clk 0:01:10] } 32. Be4 { [%clk 0:01:58] } 32... Rxb4 { [%clk 0:01:07] } 33. f3 { [%clk 0:01:58] } 33... Rb2 { [%clk 0:01:04] } 34. d4 { [%clk 0:01:55] } 34... Nc4 { [%clk 0:00:59] } 35. d5 { [%clk 0:01:53] } 35... Nd6 { [%clk 0:00:57] } 36. Rc1 { [%clk 0:01:49] } 36... g5 { [%clk 0:00:48] } 37. Rc6 { [%clk 0:01:48] } 37... Nxe4 { [%clk 0:00:42] } 38. fxe4 { [%clk 0:01:47] } 38... Kg7 { [%clk 0:00:40] } 39. d6 { [%clk 0:01:46] } 39... Rd2 { [%clk 0:00:39] } 40. e5 { [%clk 0:01:45] } 40... Kf8 { [%clk 0:00:26] } 41. Rc7 { [%clk 0:01:43] } 41... Ke8 { [%clk 0:00:24] } 42. Kf1 { [%clk 0:01:38] } 42... h5 { [%clk 0:00:13] } 43. Rxf7 { [%clk 0:01:37] } 43... Kxf7 { [%clk 0:00:09] } 44. g3 { [%clk 0:01:36] } 44... Ke6 { [%clk 0:00:08] } 45. d7 { [%clk 0:01:35] } 45... Rxd7 { [%clk 0:00:06] } 46. h3 { [%clk 0:01:34] } 46... Kxe5 { [%clk 0:00:05] } 47. Kf2 { [%clk 0:01:34] } 47... Rd3 { [%clk 0:00:04] } 48. Kg2 { [%clk 0:01:31] } 48... Ra3 { [%clk 0:00:03] } 49. Kf2 { [%clk 0:01:30] } 49... Ke4 { [%clk 0:00:03] } 50. Kg2 { [%clk 0:01:30] } 50... Ra2+ { [%clk 0:00:02] } 51. Kf1 { [%clk 0:01:28] } 51... Ke3 { [%clk 0:00:01] } 52. Ke1 { [%clk 0:01:27] } 52... Ra1# { [%clk 0:00:00] } 0-1

and to run the visitor:

vis = TestVisitor(game)
game.accept(vis)

[guidopetri]

Additionally, running the test after allowing the visitor to run over the chess game yields the incorrect behavior.

[zhelyabuzhsky]

Can you write test (pull request) to reproduce this bug?

[guidopetri]

Created the test in #22 . Now to find out how to fix this bug...

[zhelyabuzhsky]

Thank you

I'll try to help you =)

[zhelyabuzhsky]

What about #23?

[guidopetri]

I am thinking it might be better to add it in get_best_move as well... I'm not sure if calling different depths gives different results there. I'll try and write a test for this too.

[zhelyabuzhsky]

Ok, I'll merge this pr. If you have a idea to improve this, I will always be glad your PR =)

[guidopetri]

Sorry, I've been busy with school :P will let you know when I've thought about this more. Thanks for merging the test :)