Introduction

Python implementation of the chess game in less than 1000 lines of code.

For engines/agents to play against check my other project chess-ai

There is also a work in progress for the cli based game (python -m chess_python.game).

Usage

pip install chess-python-pip

#examples/example_01.py
from chess_python.chess import Chess
chess = Chess(fen="rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 0") # Default
print(chess)
print(chess.legal_moves())
chess.move("e2e4")
print(chess)

outputs:

Player to move: White
Move count: 0
8 || r | n | b | q | k | b | n | r |
-----------------------------------
7 || p | p | p | p | p | p | p | p |
-----------------------------------
6 ||   |   |   |   |   |   |   |   |
-----------------------------------
5 ||   |   |   |   |   |   |   |   |
-----------------------------------
4 ||   |   |   |   |   |   |   |   |
-----------------------------------
3 ||   |   |   |   |   |   |   |   |
-----------------------------------
2 || P | P | P | P | P | P | P | P |
-----------------------------------
1 || R | N | B | Q | K | B | N | R |
-----------------------------------
-----------------------------------
  || a | b | c | d | e | f | g | h

['b1a3', 'b1c3', 'g1f3', 'g1h3', 'a2a3', 'a2a4', 'b2b3', 'b2b4', 'c2c3', 'c2c4', 'd2d3', 'd2d4', 'e2e3', 'e2e4', 'f2f3', 'f2f4', 'g2g3', 'g2g4', 'h2h3', 'h2h4']
Player to move: Black
Move count: 1
8 || r | n | b | q | k | b | n | r |
-----------------------------------
7 || p | p | p | p | p | p | p | p |
-----------------------------------
6 ||   |   |   |   |   |   |   |   |
-----------------------------------
5 ||   |   |   |   |   |   |   |   |
-----------------------------------
4 ||   |   |   |   | P |   |   |   |
-----------------------------------
3 ||   |   |   |   |   |   |   |   |
-----------------------------------
2 || P | P | P | P |   | P | P | P |
-----------------------------------
1 || R | N | B | Q | K | B | N | R |
-----------------------------------
-----------------------------------
  || a | b | c | d | e | f | g | h

DEV section:

Performance tracking

Using python3 -m cProfile -o prof.txt tree.py -h for profiling and snakeviz prof.txt to visualize.

Perft(3) initial position rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1 (8902 positions, reference time 0.05 s with python-chess:

• 1.25 s (first)
• 0.79 s (moving get_positions_attacking_pieces to optimizer initialization and update, not when checking if move is legal)
• 0.70 s removing np.array from list of moves (does not make a lot of sense)
• 0.55 s removing more np.arrays
• 0.51 s removing np.unravel
• 0.47 s using only lists in get_allowed_moves_by_piece
• 0.39 s with revamped get_index_trajectory
• 0.35 with custom deepcopy
• 0.18 s with using list for board instead of numpy!!
• 0.15 s further tweaks

TODO:

• Include tests for perft in different positions
• Include utils if there is a mismatch in positions tree with reference implementation
• Improve performance:
  • Keep track of index where there are pieces in optimizer level
  • Move and unmake move (implenting unmake also requires using a copy of the board, which performance wise does not improve anything)
• Improve overall code quality (clarity, choose right data structure for the job):
  • Public vs private functions
  • cyclomatic complexity in is_legal_move
• Automate release with github action to pip
• Explore pypy
• Explore deepcopy for tree generation, it takes a lot of time (only copying board improves performance)
• Simplify castling