Basics

[Henk]

No with user interface I meant the driver part of my engine that prints the move. If it does nor use the driver it plays the internal move when it uses it's own GUI.

[Henk]

(Changed it.) By the way at level 1 Skipper does not see stalemate. (It also should not add h3h4)

Code: Select all

Depth  Value   Time(seconds)   Nodes
  3        M3       0.06         676   f1=R h2h3 f1h1 h3h4 
  2    9.4576       0.06         313   f1=R h2h3 
  1   11.9576       0.06          71   f1=Q 

[mar]

By the way at level 1 Skipper does not see stalemate.

That's ok, you don't have to detect stalemate in qs.

[Daniel Anulliero]

(Changed it.) By the way at level 1 Skipper does not see stalemate. (It also should not add h3h4)

Code: Select all

Depth  Value   Time(seconds)   Nodes
  3        M3       0.06         676   f1=R h2h3 f1h1 h3h4 
  2    9.4576       0.06         313   f1=R h2h3 
  1   11.9576       0.06          71   f1=Q 

Funny pv at depth 3
Rh1+ and Kh4 illegal move

[Henk]

Actually I think KRKN end game occurs more frequently than KRK or otherwise KRKNP. See Skipper often playing that endgame and always losing.

Welllll. ... KRKN is usualy à draw ...
I have a scoop for you ... you have bugs somewhere
Need to fix them...
Where can I see skipper playing ? Any download?
Personnaly I'll plan to release Isa soon , sometimes , the Testers or the "magicians of compiler" can help to fix tons of things .
You should release Skipper too

Yes terrible bug(s). Skipper played in this position Rf2: ???? Don't know what has happened. Can't reproduce. I remember Skipper giving away a bishop a few days ago but I thought it had to do with a too large passed pawn bonus.


[d] 6k1/5p1p/6p1/1p6/2p2P1B/7P/4rP2/1K6 b - - 3 41


See not reproducible:

Code: Select all

Depth  Value   Time(seconds)   Nodes
 13    5.0408       0.37      442114   b5b4 h4f6 e2f2 f6e5 c4c3 h3h4 c3c2 b1b2 b4b3 e5b8 g8g7 b8d6 f2e2 
 12    5.0856       0.29      329596   b5b4 h4f6 e2f2 f6e5 c4c3 h3h4 c3c2 b1c1 b4b3 c1b2 f2d2 b2c1 
 11    5.0408       0.23      253375   b5b4 h4f6 e2f2 f6e5 c4c3 h3h4 c3c2 b1b2 b4b3 e5d6 f2e2 
 10    4.9896       0.17      158083   b5b4 h4f6 e2f2 f6e5 c4c3 b1a1 c3c2 a1b2 b4b3 h3h4 
  9    4.4904       0.09       57705   b5b4 b1a1 b4b3 h4f6 e2a2 a1b1 a2f2 f6e5 f7f5 
  8    4.3560       0.07       29370   b5b4 b1a1 b4b3 h4g3 e2a2 a1b1 c4c3 h3h4 
  7    4.2344       0.06       14085   b5b4 f2f3 c4c3 h4f6 c3c2 b1b2 b4b3 
  6    4.2344       0.05       10593   b5b4 f2f3 c4c3 h4f6 c3c2 b1b2 b4b3 
  5    4.1704       0.05        4164   b5b4 f2f3 c4c3 h4f6 c3c2 b1b2 
  4    3.7096       0.04        1713   b5b4 f2f3 c4c3 h4f6 
  3    3.7416       0.04         749   b5b4 f2f3 c4c3 
  2    3.3896       0.04         262   b5b4 f2f3 
  1    3.3960       0.04          67   b5b4 

[Henk]

KRK ? It cannot even win KQK. Also 6. .. Bc3: hopeless bad move.

[pgn]
[Event "Computer Chess Game"]
[Site "HP"]
[Date "2016.03.16"]
[Round "-"]
[White "Fairy-Max 4.8S"]
[Black "SkipperWinb"]
[Result "1/2-1/2"]
[TimeControl "120"]
[Annotator "1. +0.13 1... +0.00"]

1. Nc3 {+0.13/8} g6 {+0.00/13 2.5} 2. d3 {+0.06/8 5} Bg7 {+0.02/14 2.4} 3.
h4 {+0.02/7 1.8} Nf6 {+0.58/13 2.3} 4. Bf4 {-0.14/8 2.6} Nh5 {+0.18/13 2.3}
5. Bd2 {-0.35/9 3} O-O {+0.61/13 2.3} 6. e3 {-0.10/8 2.9} Bxc3
{+0.43/9 2.2} 7. bxc3 {+0.25/10 7} b6 {-0.09/10 2.2} 8. Be2 {+0.43/10 1.4}
Nf6 {-0.06/14 2.1} 9. Nf3 {+0.37/10 4} Bb7 {-0.07/15 2.1} 10. O-O
{+0.39/10 3} Nc6 {-0.18/14 2.0} 11. a3 {+0.33/10 1.9} d5 {-0.21/13 2.0} 12.
c4 {+0.34/10 2.1} dxc4 {-0.14/14 2.0} 13. dxc4 {+0.09/10 4} Ne4
{-0.15/14 2.0} 14. Be1 {+0.21/9 1.0} Qxd1 {-0.20/13 1.9} 15. Rxd1
{+0.18/10 1.6} Rad8 {-0.18/13 1.9} 16. Rd5 {+0.23/9 2.1} Ba6 {-0.03/13 1.8}
17. Bd3 {+0.13/10 1.4} Nc5 {-0.14/14 1.8} 18. Rxd8 {-0.10/11 1.7} Nxd8
{-0.22/14 1.8} 19. Be2 {+0.00/10 0.9} Bb7 {-0.17/13 1.7} 20. Nd4
{+0.13/9 1.7} a5 {-0.16/14 1.7} 21. f4 {+0.15/9 1.0} Ne4 {-0.14/15 2.1} 22.
Nb5 {+0.18/10 1.4} Ne6 {-0.16/14 2.0} 23. Nc3 {+0.34/9 1.4} Rd8
{-0.14/14 2.0} 24. Bf3 {+0.23/10 1.8} N6c5 {-0.12/13 1.9} 25. Nxe4
{+0.01/10 1.2} Bxe4 {+0.47/14 2.0} 26. Bxe4 {-0.26/10 0.8} Nxe4
{+0.51/16 1.9} 27. f5 {-0.37/10 0.8} gxf5 {+0.65/14 1.7} 28. Rxf5
{-0.54/11 1.0} Rd1 {+1.30/16 1.8} 29. Rf1 {-0.72/12 1.7} Ra1 {+1.33/17 1.6}
30. h5 {-0.58/11 1.1} Rxa3 {+1.40/12 1.6} 31. h6 {-1.01/11 1.9} Rxe3
{+2.00/14 1.6} 32. Bh4 {-1.06/11 1.4} f6 {+2.00/12 1.5} 33. Rf4
{-1.02/10 0.9} Nd6 {+2.34/13 1.6} 34. Rg4+ {-1.14/10 0.9} Kh8
{+2.35/13 1.5} 35. Bf2 {-1.24/11 1.2} Re4 {+2.47/15 1.5} 36. Rxe4
{-1.71/14 1.1} Nxe4 {+2.52/15 1.5} 37. Bd4 {-1.86/13 4} e5 {+2.90/17 1.3}
38. Bb2 {-1.74/12 1.7} c5 {+2.95/15 1.3} 39. Kf1 {-1.61/12 1.0} Nd2+
{+3.27/17 1.3} 40. Ke2 {-1.71/13 0.9} Nxc4 {+3.46/17 1.3} 41. Bc3
{-2.26/12 0.7} a4 {+3.46/15 0.9} 42. Kd3 {-2.04/12 0.7} b5 {+3.70/15 0.9}
43. Ke4 {-2.47/12 0.9} a3 {+3.69/13 0.8} 44. Kd5 {-2.60/12 1.0} a2
{+5.43/17 0.9} 45. Kxc5 {-1.84/12 0.8} Ne3 {+5.69/16 0.9} 46. Kxb5
{-1.94/12 0.6} Nxc2 {+4.67/19 0.8} 47. Kc4 {-2.00/13 0.5} a1=Q
{+4.68/20 0.7} 48. Bxa1 {-3.14/15 0.9} Nxa1 {+4.68/18 0.7} 49. Kc3
{-3.62/16 0.8} Kg8 {+4.14/18 0.7} 50. Kb2 {-3.63/16 0.6} e4 {+4.46/21 0.7}
51. g4 {-4.18/17 0.8} e3 {+4.81/14 0.7} 52. Kc3 {-3.66/17 0.5} e2
{+4.81/12 0.7} 53. Kd2 {-3.62/17 0.7} e1=Q+ {+4.85/17 0.7} 54. Kxe1
{-3.65/17 0.6} Nc2+ {+4.44/20 0.7} 55. Kd2 {-3.47/16 0.4} Nb4
{+4.46/20 0.7} 56. Kc3 {-3.57/16 0.6} Nc6 {+4.46/20 0.7} 57. Kc4
{-4.30/16 1.0} Kf7 {+4.48/20 0.7} 58. Kd5 {-4.24/15 0.6} Ne7+
{+4.48/20 0.6} 59. Kd6 {-4.36/15 0.4} Ng8 {+4.62/16 0.6} 60. Kd5
{-4.63/15 0.5} Nxh6 {+4.72/19 0.6} 61. Kd4 {-4.73/15 0.5} Nxg4
{+4.73/18 0.6} 62. Ke4 {-5.32/15 0.5} Kg6 {+4.89/17 0.6} 63. Kf3
{-6.06/16 0.5} h5 {+5.04/18 0.6} 64. Kf4 {-6.78/16 0.4} f5 {+5.65/21 0.6}
65. Kf3 {-10.29/16 0.8} Kh6 {+5.65/20 0.6} 66. Kf4 {-8.79/18 0.6} h4
{+8.20/16 0.6} 67. Kxf5 {-8.81/18 0.3} h3 {+8.20/14 0.6} 68. Kxg4
{-8.80/17 0.5} h2 {+8.17/12 0.5} 69. Kf4 {-8.80/16 0.5} h1=Q {+8.17/10 0.5}
70. Ke5 {-79.93/16 0.5} Qc6 {+8.17/9 0.5} 71. Kd4 {-79.93/15 0.5} Kg5
{+7.21/7 0.5} 72. Ke3 {-79.94/15 0.3} Qh1 {+8.11/9 0.5} 73. Kd4
{-8.82/13 0.4} Kg4 {+7.21/8 0.4} 74. Kc5 {-8.85/14 0.4} Kf3 {+7.22/10 0.4}
75. Kd6 {-8.81/12 0.3} Qh2+ {+7.22/10 0.4} 76. Kd7 {-8.83/13 0.3} Qf4
{+7.22/12 0.4} 77. Kc6 {-8.85/13 0.3} Qc4+ {+7.22/12 0.4} 78. Kb6
{-79.92/14 0.5} Qd4+ {+7.22/12 0.4} 79. Kb5 {-79.89/13 0.3} Ke4
{+7.22/13 0.4} 80. Kc6 {-79.94/14 0.3} Qc4+ {+7.22/13 0.4} 81. Kb7
{-79.95/14 0.3} Qf7+ {+7.22/14 0.4} 82. Kc6 {-79.94/14 0.3} Kf3
{+7.22/14 0.4} 83. Kd6 {-79.91/14 0.3} Qf4+ {+7.22/13 0.4} 84. Kc6
{-0.01/28 0.2} Ke4 {+7.22/14 0.4} 85. Kc5 {-79.94/14 0.2} Kf3
{+7.22/15 0.4} 86. Kc6 {-0.01/28 0.2}
{XBoard adjudication: repetition draw} 1/2-1/2
[/pgn]

[flok]

By the way at level 1 Skipper does not see stalemate.

That's ok, you don't have to detect stalemate in qs.

Why not?

[mar]

That's ok, you don't have to detect stalemate in qs.

Why not?

They are rare and you can detect stalemates in main search easily (essentially for free) - this really should do.

Now that I think about it, since you already compute mobility in eval, detecting stalemates might be relatively easy, if there is no safe mobility for the king and all other (own) pieces have zero mobility, it should be a stalemate.
Special care would be needed for pawns though, should be trivial for a bitboard engine.
So yes, why not... I still doubt it'd be worth the effort.
EDIT: this wouldn't detect all stalemates but should cover the most obvious cases

[mar]

..continuing (thanks for the 15m limit:)
So you could either simply return 0 in eval or pass the information to qs and prune.
What seems more tempting is to detect trivial checkmates in eval, I think Ed described queen contact checkmates in his technical info on Rebel.
The problem is that those ideas will tax your eval and they have to be worth the additional effort (=happen often enough) or you might lose a bit of elo.
Of course, they might be useful for analysis.
In general, they also add complexity and may introduce new bugs, so I'd be careful.

[Stan Arts]

Why not?

You probably are not going to try to generate all moves there so when you are not in check how would you know you have no moves?

One problem with that is recognising for example desperado eternal rook type situations rather than returning a won/lost score from qsearch. But most other stalemate situations are handled easily by normal search and eternal rook positions are ofcourse extremely rare. (But it would be nice to recognise them rather than going for some other losing or not winning ending. When stumbling into one ofcourse you probably get lucky and the endless checks are better than whatever else is at the horizon.)