Complicating code in C#

[Henk]

Yes. But I already have eliminated most of them. Only few left.
I have class ChessPosition and a class ChessBoard. I currently store fifty move count in ChessPosition.
So it makes a copy of ChessPosition as well as a copy of ChessBoard when making a move.
That's a bit too much. Also I store GamePhase in ChessPosition. ChessPosition has property ChessBoard.
So maybe I have to eliminate class ChessPosition as well. Put all necessary data in ChessBoard and nothing extra.

But I suspect that Stockfish is doing some clever tricks in check detection/evasion or move generation.

[Sven]

It has already been stated a few times: making movegen faster (or check detection, same topic) can't make the search faster by an order of magnitude because the search spends only a small amount of time in movegen, usually between 10% and 30%, so that even making the whole movegen three times as fast as before would only result in a search speedup between 7% and 25%.

Having a position class and a board class sounds ok. Copy&make is not unusual but it can be time consuming if you copy large and/or very complex data. You know that of course ...

If your data structures are not the reason for low speed then it could also be data flow and algorithms.

[Henk]

I also store moveNumber in ChessPosition. That is last number of the fen.
Don't know where to put it. Maybe in Game.

I only use it to check if position.ConvertToFen() == fen
But if position does not store this moveNumber anymore then that check fails.
So can't do an easy string compare,

[Henk]

Bit expensive to copy whole array for changing only one bit.
This happens when doing a null move.
It is part of copy move strategy. Instead of do undo.

Code: Select all

  
        public IChessBoard ChangePlayer()
        {
            ulong[] cp = new ulong[NBITBOARDS];
            Pieces.CopyTo(cp, 0);
            cp[OTHER] ^= CUR_PLAYER;
            var result = new ChessBoardBits(cp);
            return result;
        }
 

[Sven]

What is cp[OTHER], what is CUR_PLAYER? Both seem to be bitboards while I would expect something like White or Black ...

Also for copy&make based null move I would expect something like this (I apologize for my C# syntax sloppiness):

Code: Select all

public IChessBoard AfterNullMove()
{
    var result = new ChessBoardBits(Pieces);
    result.Player = Opponent(result.player);
    // and do possible other updates, e.g. reset ep target sqr and 50-moves counter
    return result;
}

which avoids copying the piece bitboards twice.

[Henk]

New record 53 moves. Previous record 32 moves. But Skipper played Petroff's defence.
[pgn]
[Event "Computer Chess Game"]
[Site "LAPTOP-1FK7MTIP"]
[Date "2021.04.27"]
[Round "-"]
[White "Stockfish 13"]
[Black "Skipper_8_6"]
[Result "1-0"]
[TimeControl "180"]
[Annotator "1. +0.19 1... -0.02"]

1. e4 {+0.19/26} e5 {-0.02/13 4} 2. Nf3 {+0.47/23 0.2} Nf6 {-0.01/13 4} 3.
Nxe5 {+0.47/22 1.8} d6 {-0.02/19 4} 4. Nf3 {+0.47/26 0.1} Nxe4 {-0.05/18 3}
5. d4 {+0.44/25 1.3} Nc6 {+0.00/16 3} 6. Bd3 {+0.62/24 3} d5 {-0.01/9 3} 7.
O-O {+0.58/25 2.1} Nb4 {+0.00/11 3} 8. Re1 {+1.22/23 3} Be7 {-0.13/11 3} 9.
Bxe4 {+1.37/26 0.1} dxe4 {-0.38/11 3} 10. Rxe4 {+1.26/25 0.4} Bf5
{-0.37/9 3} 11. Re2 {+1.42/24 1.2} c5 {-0.34/10 3} 12. Bg5 {+2.17/25 7} f6
{-0.12/13 2.9} 13. Bf4 {+2.21/24 0.1} cxd4 {+0.33/6 2.9} 14. Nxd4
{+2.16/24 2.8} Nc6 {-0.28/9 2.8} 15. Rd2 {+2.07/24 0.2} Nxd4 {-0.29/10 2.8}
16. Rxd4 {+2.12/24 0.2} Qc8 {-0.28/9 2.7} 17. Nc3 {+2.10/23 0.2} O-O
{-0.29/14 2.7} 18. Nd5 {+2.27/23 2.4} Bc5 {-0.08/7 2.6} 19. b4
{+3.33/25 1.8} Bxd4 {+0.09/19 2.6} 20. Ne7+ {+3.94/24 0.4} Kh8
{+0.05/17 2.5} 21. Nxc8 {+3.71/26 0.7} Bxa1 {-0.12/14 2.5} 22. Nd6
{+3.78/25 0.5} Be5 {-0.07/12 2.4} 23. Bxe5 {+3.71/24 1.9} fxe5
{-0.10/10 2.4} 24. Qd5 {+4.13/23 1.5} Bxc2 {+0.40/12 2.3} 25. Nf7+
{+4.77/24 2.7} Rxf7 {+0.00/677 4} 26. Qxf7 {+4.86/26 0.1} Rd8 {+0.00/740 4}
27. h3 {+5.00/27 0.2} Be4 {-0.58/11 2.2} 28. Qe7 {+5.01/27 1.6} Rd1+
{-1.50/20 2.2} 29. Kh2 {+5.00/24 0.5} h6 {-1.85/11 2.1} 30. Qxe5
{+5.02/27 0.9} Bd5 {-2.15/11 2.1} 31. Qb8+ {+5.56/22 2.4} Kh7
{-2.35/11 2.0} 32. Qxa7 {+5.38/24 5} g6 {-2.27/12 2.0} 33. Qc5 {+6.48/22 5}
Rd2 {-2.76/9 1.9} 34. a4 {+6.71/21 1.7} Bf7 {-2.86/11 1.9} 35. Qc7
{+7.78/21 2.8} Rxf2 {-3.27/13 1.9} 36. Qxb7 {+8.78/22 4} Kg8 {-5.28/15 1.9}
37. b5 {+9.01/20 0.4} Rd2 {-5.23/9 1.8} 38. Qc8+ {+9.64/21 2.6} Kh7
{-6.26/40 1.9} 39. Qc7 {+9.71/23 3} Kg8 {-5.86/25 1.8} 40. Qc3
{+10.09/20 0.8} Re2 {-6.25/9 1.7} 41. b6 {+10.99/19 2.9} Bd5 {-7.01/8 1.7}
42. Qc8+ {+19.27/21 2.2} Kg7 {-9.24/12 1.6} 43. Qd7+ {+24.25/23 0.8} Bf7
{-12.15/12 1.6} 44. b7 {+31.90/21 0.8} Re8 {-12.15/10 1.6} 45. a5
{+72.65/24 1.9} g5 {-12.15/8 1.5} 46. a6 {+99.77/25 2.2} g4 {-12.17/6 1.5}
47. Qxg4+ {+99.83/36 0.6} Kh8 {-16.44/7 1.5} 48. a7 {+99.87/59 2.2} Kh7
{-18.27/10 1.5} 49. b8=Q {+99.89/63 2.2} Rxb8 {-22.10/7 1.4} 50. axb8=Q
{+99.93/245 1.4} h5 {-327.60/238 2.0} 51. Qg5 {+99.95/245 0.1} h4
{-327.62/265 2.0} 52. Qf8 {+99.97/245 0.8} Be6 {-327.64/304 2.1} 53. Qgh6#
{+99.99/245 0.8}
{Xboard adjudication: Checkmate} 1-0
[/pgn]

[Henk]

What is cp[OTHER], what is CUR_PLAYER? Both seem to be bitboards while I would expect something like White or Black ...

Also for copy&make based null move I would expect something like this (I apologize for my C# syntax sloppiness):

Code: Select all

public IChessBoard AfterNullMove()
{
    var result = new ChessBoardBits(Pieces);
    result.Player = Opponent(result.player);
    // and do possible other updates, e.g. reset ep target sqr and 50-moves counter
    return result;
}

which avoids copying the piece bitboards twice.

I don't know (yet) if it is copying twice. I thought it would only assign a reference.
But not sure. I can use your code example to see if that would be faster (if it works).
I don't think this.pieces = pieces copies the content of the array.

Code: Select all

        readonly ulong[] pieces;
        private ChessBoardBits(ulong[] pieces)
        {
            this.pieces = pieces;
        }
 

By the way:

Code: Select all

   public const ulong CUR_PLAYER = ((UInt64)1) << 4;

And OTHER points to extra bitboard where sign (of current player = side to move) , en passant, castling rights etc is stored.

Code: Select all

        public const int WHITE_PIECES = (int)King + 1;
        public const int BLACK_PIECES = (int)King + 2;
        public const int OTHER = (int)King + 3;
        const int NBITBOARDS = 9;

[Henk]

What is cp[OTHER], what is CUR_PLAYER? Both seem to be bitboards while I would expect something like White or Black ...

Also for copy&make based null move I would expect something like this (I apologize for my C# syntax sloppiness):

Code: Select all

public IChessBoard AfterNullMove()
{
    var result = new ChessBoardBits(Pieces);
    result.Player = Opponent(result.player);
    // and do possible other updates, e.g. reset ep target sqr and 50-moves counter
    return result;
}

which avoids copying the piece bitboards twice.

No does not work. As expected. Constructor does not make copy of pieces.
this.pieces = pieces; does not copy array.
To make it work I must make copy of pieces explicitly.

Code: Select all

public IChessBoard ChangePlayer()
{
  var cp = new ChessBoardBits(Pieces); // does not work 
   cp.Pieces[OTHER] ^= CUR_PLAYER;
   return cp;
}

This code works. Performance similar.
Better do a CopyTo in constructor.

Code: Select all

public IChessBoard ChangePlayer()
{
            ulong[] cpPieces = new ulong[NBITBOARDS];
            Pieces.CopyTo(cpPieces, 0);
            var cp = new ChessBoardBits(cpPieces);
            cp.Pieces[OTHER] ^= CUR_PLAYER;
            return cp;
 }
 

[Henk]

Another win against fairy-max. They are still scarce so I post them.
Same story: opponent castling long and a win before end game starts.
Because when in end game Skipper always loses somehow.

[pgn]
[Event "Computer Chess Game"]
[Site "LAPTOP-1FK7MTIP"]
[Date "2021.04.29"]
[Round "-"]
[White "Skipper_8_7"]
[Black "Fairy-Max 4.8V"]
[Result "1-0"]
[TimeControl "180"]
[Annotator "1. +0.37 3... +0.26"]

1. d4 {+0.37/14} d5 2. Nc3 {+0.20/13 4} Nf6 3. Bf4 {+0.21/13 3} c6
{+0.26/7 6} 4. e3 {+0.21/12 3} Be6 {+0.11/7 4} 5. Bd3 {+0.35/12 3} Nbd7
{+0.08/7 3} 6. Nge2 {+0.28/12 3} h6 {+0.04/7 2.3} 7. h4 {+0.31/12 3} Nh5
{+0.01/7 2.7} 8. Bh2 {+0.37/13 3} Qb6 {+0.00/8 6} 9. Rb1 {+0.71/11 3} g5
{+0.04/7 2.7} 10. hxg5 {+0.78/13 3} hxg5 {-0.14/8 2.1} 11. O-O {+0.78/12 3}
O-O-O {+0.18/7 3} 12. b4 {+1.21/11 2.9} Re8 {-0.07/9 4} 13. b5
{+1.01/12 2.9} Qa5 {-0.01/7 3} 14. bxc6 {+0.85/11 2.8} bxc6 {-0.05/8 2.8}
15. e4 {+0.91/10 2.8} Nhf6 {+0.02/8 5} 16. e5 {+0.92/11 2.7} Ng4
{-0.10/9 4} 17. Bg3 {+0.78/12 2.7} f5 {+0.04/8 17} 18. exf6 {+0.65/12 2.6}
Ngxf6 {-0.04/8 3} 19. Re1 {+0.71/10 2.6} Nh5 {+0.04/8 1.8} 20. Bg6
{+0.72/10 2.5} Rd8 {+0.03/9 5} 21. Bh2 {+0.27/11 2.5} Nhf6 {-0.06/8 2.1}
22. Bd3 {+0.53/10 2.4} Bg4 {-0.07/6 1.3} 23. f3 {+0.88/10 2.4} Be6
{-0.66/7 1.9} 24. Qd2 {+0.96/11 2.3} Qa3 {-0.77/6 1.7} 25. Nf4
{+1.67/11 2.3} Bg8 {-1.11/7 1.5} 26. Ne6 {+2.72/12 2.2} Ne4 {-2.05/7 5} 27.
Nxe4 {+4.01/11 2.2} Rxh2 {-3.05/6 3} 28. Qc3 {+6.96/12 2.1} Qa4
{-7.80/8 1.6} 29. Bb5 {+10.16/12 2.1} Qxb5 {-9.77/8 1.8} 30. Rxb5
{+14.51/14 2.1} Ne5 {-11.48/8 1.6} 31. dxe5 {+14.21/13 2.0} Kd7
{-13.76/9 2.4} 32. Nxf8+ {+14.49/12 2.0} Ke8 {-13.30/9 2.0} 33. Qxc6+
{+14.78/12 1.9} Kxf8 {-14.79/10 1.9} 34. Qc7 {+14.96/11 1.9} Re8
{-11.85/9 1.5} 35. Kxh2 {+14.79/12 1.9} dxe4 {-12.23/9 0.9} 36. fxe4
{+15.56/13 1.8} Bxa2 {-12.91/9 1.5} 37. c4 {+17.37/13 1.8} Kg7
{-13.49/8 1.3} 38. Ra1 {+17.50/12 1.8} a6 {-13.94/10 0.9} 39. Rb6
{+18.53/14 1.7} Bxc4 {-14.24/10 2.1} 40. Qxc4 {+18.74/13 1.7} e6
{-19.27/10 1.9} 41. Raxa6 {+19.62/13 1.7} Rh8+ {-79.94/9 0.9} 42. Kg3
{+20.60/14 1.6} Rf8 {-79.96/12 1.0} 43. Rxe6 {+327.59/14 1.6} Rf7
{-79.97/18 0.9} 44. Rg6+ {+327.61/15 1.6} Kh8 {-79.98/28 0.2} 45. Qc8+
{+327.63/16 1.5} Kh7 {-79.99/28 0.2} 46. Qh3# {+327.65/17 1.5}
{Xboard adjudication: Checkmate} 1-0
[/pgn]

[Henk]

Another win but in the mean time already lost several games.
But looks like wins getting less scarce now.

[pgn]
Event "Computer Chess Game"]
[Site "LAPTOP-1FK7MTIP"]
[Date "2021.04.29"]
[Round "-"]
[White "Skipper_8_7"]
[Black "Fairy-Max 4.8V"]
[Result "1-0"]
[TimeControl "180"]
[Annotator "1. +0.37 2... +0.20"]

1. d4 {+0.37/14} Nf6 2. Nc3 {+0.00/12 4} Nc6 {+0.20/8 4} 3. d5 {+0.57/13 3}
Nb8 {-0.08/8 6} 4. Nf3 {+0.59/12 3} d6 {+0.12/8 7} 5. e4 {+0.80/13 3} c6
{+0.14/7 2.3} 6. Be3 {+0.72/12 3} Nbd7 {+0.14/8 11} 7. Nd4 {+0.55/11 3} Ne5
{+0.05/8 5} 8. h3 {+0.68/11 3} cxd5 {+0.02/8 5} 9. exd5 {+0.84/12 3} Bd7
{-0.06/7 2.3} 10. f4 {+0.81/12 3} Ng6 {-0.21/8 2.6} 11. Bc4 {+0.68/10 3}
Qb6 {+0.10/7 3} 12. Qd3 {+1.14/9 2.9} Qxb2 {+0.02/6 2.0} 13. Rb1
{+1.15/11 2.9} Qa3 {-0.28/8 2.4} 14. Rxb7 {+1.15/9 2.8} a6 {-0.35/7 3} 15.
O-O {+2.01/9 2.8} Bc8 {-0.08/7 5} 16. Rb6 {+2.20/9 2.7} Rg8 {-0.01/6 1.8}
17. Rfb1 {+2.39/8 2.7} Rh8 {-0.23/6 1.6} 18. R1b3 {+3.69/10 2.6} Qc5
{-1.79/9 4} 19. Ne6 {+3.64/11 2.6} Nxf4 {-2.24/8 2.1} 20. Nxf4
{+4.84/11 2.5} Qc7 {-2.40/7 2.1} 21. Bxa6 {+5.00/9 2.5} Bd7 {-2.54/7 1.5}
22. Bb7 {+5.46/10 2.4} Rb8 {-2.63/7 1.8} 23. Bd4 {+4.59/10 2.4} Qd8
{-2.94/7 4} 24. Qe3 {+5.06/10 2.3} Qc7 {-3.06/7 2.6} 25. Nb5 {+5.11/10 2.3}
Qxb6 {-4.59/8 1.5} 26. Bxb6 {+5.59/12 2.2} Bxb5 {-5.46/9 1.7} 27. Bc6+
{+6.92/12 2.2} Bxc6 {-5.01/10 1.7} 28. dxc6 {+6.92/12 2.1} h5 {-5.16/9 4}
29. c7 {+6.37/11 2.1} Rc8 {-5.64/9 1.9} 30. h4 {+6.22/10 2.1} e5
{-5.37/8 1.9} 31. Qe2 {+6.16/10 2.0} g6 {-5.72/8 3} 32. Rd3 {+6.21/8 2.0}
Nd7 {-5.16/9 1.8} 33. Ba5 {+5.85/10 1.9} Be7 {-5.48/8 1.9} 34. Nd5
{+6.42/11 1.9} Nc5 {-5.96/7 1.2} 35. Rh3 {+6.63/10 1.9} f6 {-6.06/8 4} 36.
Rg3 {+7.00/9 1.8} f5 {-6.65/7 1.0} 37. Rxg6 {+7.49/11 1.8} Ra8
{-6.87/7 1.2} 38. Nxe7 {+7.79/8 1.7} Kxe7 {-7.58/9 1.8} 39. Bb4
{+8.71/10 1.7} Kd7 {-7.65/7 1.2} 40. Qd1 {+8.97/10 1.7} Nb7 {-8.44/8 1.2}
41. Bxd6 {+10.09/12 1.7} Rae8 {-9.56/8 2.1} 42. Bxe5+ {+13.23/12 1.6} Kc8
{-17.60/10 1.3} 43. Rg7 {+15.80/11 1.6} Rd8 {-18.49/11 1.1} 44. cxd8=Q+
{+21.22/12 1.6} Rxd8 {-79.96/12 1.0} 45. Qf3 {+24.85/12 1.5} Rd1+
{-79.97/18 0.9} 46. Qxd1 {+25.54/12 1.5} Nc5 {-79.98/28 0.2} 47. Rc7+
{+327.63/12 1.5} Kb8 {-79.99/28 0.2} 48. Qd8# {+327.65/13 1.4}
{Xboard adjudication: Checkmate} 1-0
[/code][/pgn]