How can I make my chess engine more efficient?

Question

First, here's some context. My chess engine uses bitboards and works like this: A single gameboard is created as a 2d array at the start of the program. Like this:

[[-4 -2 -3 -5 -6 -3 -2 -4]
 [-1 -1 -1 -1 -1 -1 -1 -1]
 [ 0  0  0  0  0  0  0  0]
 [ 0  0  0  0  0  0  0  0]
 [ 0  0  0  0  0  0  0  0]
 [ 0  0  0  0  0  0  0  0]
 [ 1  1  1  1  1  1  1  1]
 [ 4  2  3  5  6  3  2  4]]

Where each number represents a piece. This 2d board is then converted to 12 bitboards, each representing a set of white or black pieces for each color. The possible legal moves are then generated using various bitboard techniques.

Now to my question: I've written the engine in Python, and so far, the search and evaluate function (a.k.a the minimax function) can only reach a ply of 4. When I try to start going to a ply of 6, the program takes about a good 45-60 seconds to calculate the best move. Any higher ply would make the computer take too long.

Is my implementation inefficient? If not, then any advice on why my engine is so slow would be appreciated. I imagined that with bitboards I could at least reach a ply of 8-12 in Python, even though it's not a very efficient language.

Edit: Also yes, I'm using alpha-beta pruning.

Answer 1

Generally, I would avoid using Python for chess programming altogether. Consider converting to Cython or better yet, use a modern system language such as Rust, Zig, etc.

Answer 2

If you are not yet using move ordering, I would strongly suggest you do so. Even some basic MVV-LVA helped increase my Python engine's speed by almost 500 percent. Here is some sample code if you want:

def OrderMoves(board, moveList):
    moveScores = []
    pieceIndexDelta = 6 * sideToMove  # given as zero or one, one for white
    for moveIndex in range(len(moveList)):
        currentMove = moveList[moveIndex]
        if move is castling:  # this is only necessary if you handle castling weirdly
            moveScores.append(0)
        elif move is promotion:
            moveScores.append(37 + promotionType)
        elif board[endSquare]:  # if move is a capture
            moveScores.append(6 * (board[endSquare] - pieceIndexDelta) -(board[startSquare] + PieceIndexDelta - 6))
        else:
            moveScores.append(0)
    return [n for _, n in sorted(zip(moveScores, moveList), reverse = True)]

This function will return the moves but sorted with good captures in front. If you have already done this I recommend you explore some more options such as null move pruning, MTD(f), iterative deepening, and quiescence search, all of which are easily accessible on Chess Programming Wiki.

Edit: for this code my piece list is as follows:

0: empty square
1: white pawn
2: white knight
3: white bishop
4: white rook
5: white queen
6: white king
7: black pawn
8: black knight
9: black bishop
10: black rook
11: black queen
12: black king