How NagaSkaki plays chess

How NagaSkaki plays chess

NagaSkaki is a young program and a lot of enhancements must still be implemented in its engine. Here is a summary of its current engine:

The chessboard is represented with bitboards.
Searching is done with an alpha beta algorithm with: extensions (check, recapture, etc.), lazy evaluation and futility pruning
It uses Null move forward pruning in the opening, middle game and ending when both sides have at least one piece (with R=2 and R=3).
NagaSkaki's positional knowledge has not been fine tuned and it seems to play rather aggressive.
It uses a small opening book that can handle transpositions.
Other features it has: Killer moves, History heuristics, quiescence search, Aspiration window and Transposition tables.

Move Generation is a bit different than the normal rotated bitboards (Actually they're not rotated at all!) and works as follows:

First Generate a bitboard of all squares occupied by pieces.
Next all the bits, except those in the direction we want to move (left, right, up and down for rook), are deleted.
Now we "shift" all the bits to the direction we want to move, e.g. for a rook moving right from the A file: The occupied board 0001 1010 becomes: 0000 1101 | 0000 0110 | 0000 0011 = 0000 1111 (I only show the occupied rank - of course the whole bitboard consists of 64 bits.)
We then XOR the bits, so that all bits where we can move are set (1111 0000).
AND the result with all bits, except those where our own pieces are, so that we only generate captures of enemy pieces and not our own.
Lastly, convert the bitboard to moves.

Here is an example of the generation of horizontal rook moves:
bitboard rightboard = (occupiedboard & rookXrays[sq][0]); //rookXrays[sq][0] is all bits to the right of rook
rightboard = rightboard<<1|rightboard<<2|rightboard<<3|rightboard<<4|rightboard<<5|rightboard<<6;
rightboard&=rookXrays[sq][0];
rightboard=(rightboard^rookXrays[sq][0]) & notfriendlyboard;
bitboard leftboard = (occupiedboard & rookXrays[sq][2]); //rookXrays[sq][2] is all bits to the left of rook
leftboard=leftboard>>1|leftboard>>2|leftboard>>3|leftboard>>4|leftboard>>5|leftboard>>6;
leftboard&=rookXrays[sq][2];
leftboard=(leftboard^rookXrays[sq][2]) & notfriendlyboard;
bitboardToMoves(rightboard|leftboard);

Last updated: 22 March 2004