chessmate 1.0.7

Framework for defining and analyzing chess engines

chessmate

chessmate is a framework built around python-chess that enables a programmatic approach to analyses and defining of chess engines.

---

Installation

Requires python 3.5 and higher

pip install chessmate

---

Usage

Engines

The basic building block of chessmate is the engine. All engines inherit from chessmate.engines.BaseEngine and obey an evaluate-move progression wherein the engine evaluates the current board state by some metric and/or algorithm and returns a single move it deems best. Mathematically, an engine can be defined as f(boardstate) = move

Examples of some simple engines included in chessmate are:

1. Random - returns a random move
2. PrioritizePawnMoves - prioritizes all moves pawn related
3. CaptureHighestValue - prioritize capturing the highest value piece available
4. ScholarsMate - obeys standard Scholar's Mate sequence and resigns if unsuccessful

Also included in the engines is the MiniMax engine, which utilizes the MiniMax algorithm to evaluate board states. The MiniMax algorithm provided comes with most bells and whistles: alpha-beta pruning, move ordering via. MVV-LVA, and a transposition table.

Since almost all chess engines can be boiled down to this basic progression, the chessmate engine schema provides a simple but powerful framework for developing and analyzing engines. More powerful chess engines (Stockfish, DeepBlue, etc) can be written to fit this schema and thus be analyzed within chessmate.

Evaluation functions

Each engine consists of an evaluation function which inherits from chessmate.analysis.EvaluationFunction. An evaluation function can be defined mathematically as f(boardstate) = evaluation where evaluation is a numeric representation of the state of the board. If we expand the definition of an engine to f(eval_function(boardstate)) = move, we can start optimizing engines via. the evaluation function.

Examples of some evaluation functions included with chessmate are:

1. StandardEvaluation - returns an evaluation based off the material difference on the board
2. PiecePositionEvaluation - returns an evaluation based off the relative position of pieces on the board

Each engine is by default configured with the StandardEvaluation function but can be mapped to any evaluation function via. the self.evaluation_function attribute

Piece values and piece value tables

Each evaluation function utilizes defined piece values and piece value tables from chessmate.constants.piece_values. Piece values provide the fundamental value of a piece on a board. By defining the value of each piece under a given condition, the evaluation function can be made to prioritize certain pieces, boardstates, or strategies.

Move ordering

chessmate engines come predefined with move-ordering capabilities defined in heuristics.py. Move ordering is defined as a heuristic function which is then incorporated into the engine. For example, adding SEE move ordering to the minimax algorithm is as simple as:

def SEE(board: chess.Board) - > List[chess.Move]:
  """ Move ordering via SEE algorithm """
  pass
  
minimax = MiniMax(color=chess.WHITE)
minimax.ordering_heuristic = SEE

---

Game simulations

Once engines are defined, one can perform analysis via the classes available in chessmate.simulations. Some example functionality includes:

Simulating a game between two engines:

from chessmate.simulations import ChessPlayground
from chessmate.engines import CaptureHighestValue, Random

# Setup simulated game between ScholarsMate engine on white and CaptureHighestValue engine on black.
simulation = ChessPlayground(ScholarsMate(), CaptureHighestValue())
simulation.play_game()

Simulating multiple games

# Setups multiple independently simulated games
simulation.play_multiple_games(1000)

One can also play directly against an engine in the IPython console:

playvs = PlayVsEngine(CaptureHighestValue())
playvs.play_game()

---

Basic analysis

To evaluate the results of a simulation:

Since the ScholarsMate engine either successfully mates or resigns, we'd expect a small percentage of games to be won by white mating and the rest black by resignation.

from chessmate.utils import display_all_results
display_all_results(simulation.all_results)

To view the difference in material across a game or games

from chessmate.utils import display_material_difference

# Use CaptureHighestvalue on white and Random engine on black
simulation = ChessPlayground(CaptureHighestValue(), Random())
simulation.play_multiple_games(10)
display_material_difference(simulation.all_material_differences, game_index=0)

To visualize the events of a game, chessmate comes with IPython functionality to display games in the console move by move

from chessmate.utils import walkthrough_pgn
walkthrough_pgn(simulation.game_pgns[0])

---

Contributing

Contributions at all levels are welcome! I'm happy to discuss with anyone the potential for contributions. Please see CONTRIBUTING.md for some general guidelines and message me with any questions!

---

Meta

Jiaming Chen – [e-mail]

Distributed under the GPL 3 (or any later version) license. See LICENSE for more information.

https://github.com/sansona/chessmate