The author demonstrates an artificial data compression technique for storing chess games based on knowledge of chess. The main idea of the approach, called “compression by prediction,” is clear--moves that are more likely according to chess theory, or classified by a higher Elo rating, or predicted by a suitable deterministic chess program have shorter codes. The construction of the program realizing this consideration is strain-forward and is based on the selection of the three best moves by a suitable deterministic chess program. The approach is connected to the well-developed idea of invertible coding using a statistical prediction method.

As a whole, this note may be treated as an example of a data-compression method for databases based upon the existence of some goodness estimation function. Unfortunately, it seems that most decompression methods of this type have complexities comparable with that of the exhaustive description of the database objects, and the alternative is direct accumulation of the corresponding probabilities (this alternative is, formally, a chess theory approach).

The paper is well written and contains interesting concrete information; it is easy to read because of the tutorial style. It may be used as a basis for student projects, especially as it contains general questions about the subject.