A collection of simple chess engines in Python, trained using datasets from Lichess
Jackson Campolattaro || PID: jackcamp
Muskaan Narula || PID: muskaannarula22
Michael Fuhrer || PID: mfuhrer
We produced several models during our development process,
all chess engines are contained in the engines directory.
The relevant engines for this project include naive_bayes and deep_learning.
Other engines include old_naive_bayes which chooses the most common move without looking
at the current state of the board,
and random_engine which chooses randomly from the pool of legal moves.
Each engine provides a class (e.g. NaiveBayesEngine) which implements the Engine interface
defined in engine.py. The classes provide the following functionality:
engine.train(filepath)Load a dataset of games, and use the data to train the model.engine.choose_move(board)Returns the model's choice of move, based on the current board state.engine.save_model(filename)Saves the current (trained) model in a file adjacent to the engine's file.engine.load_model(filename)Loads a previously saved model from a file adjacent to the engine's file.
For each engine, there is also a train.py, running this
invokes the engine's train method and saves the new model.
We also provide *program.py for each engine.
Running this starts an engine which speaks the UCI standard,
this allows us to use the engines with different interfaces.
uci.py is the adapter which enables the use of different Engine objects
with the UCI standard.
To try out one of the engines, we recommend running uci_game.py
with python 3.8. (With a modern IDE, dependencies should be automatically read
from requirements.txt and installed in your venv)
uci_game.py provides an ascii chess board with which you can play against
an engine that conforms to our interface.
Engines can be swapped by changing the first line of main() to load
a different engine.
Play is done by sending commands in the UCI format, for example, "e2e3" moves the pawn at column E forward by 1 square.