COMPR[CH]ESS

An innovative approach to lossless chess game compression, seamlessly blending chess engines, machine learning, and Huffman Coding.

DEMO VISUALIZATION

HOW IT WORKS

This program utilizes Huffman Coding to encode/decode chess games. Huffman Coding is a technique where more probable characters or strings of characters are assigned lower-length, unique bitstrings. In this context, more probable chess moves are given shorter bitstrings.

A machine learning model is used to predict the next move in a chess game. Models are trained on a dataset of thousands of chess games from lichess, filtered to only include games from players with average ratings inside a certain range. As a result, the model is able to accurately predict the next move from a given position based on the player's rating.

From probabilities given by the model, Huffman Coding is used to adaptively encode/decode chess games into binary strings. Due to the model being trained on games from players with similar ratings, the codes generated accurately reflect the probability of a move being played in a given position.