Automated Chess Tutor

Abstract

While recently the strength of chess-playing programs has grown immensely, their capability of explaining in human understandable terms why some moves are good or bad has enjoyed little attention. Progress towards programs with an ability to provide intelligent commentary on chess games, either played by a program or by a human, has been negligible in comparison with the progress concerning playing strength. The typical style of a program’s “comments” (in terms of the best variations and their numerical scores) is of little use to a human who wants to learn important concepts behind the variations.

In this paper, we present some core mechanisms for automated commenting in terms of relevant goals to be achieved or preserved in a given position. By combining these mechanisms with an actual chess engine we were able to transform this engine into a chess tutor/annotator that is capable of generating rather intelligent commentary. The main advantages of our work over related approaches are: (a) it has the ability to act as a tutor for the whole game of chess, and (b) it has a relatively solid chess understanding and is thus able to adequately comment on positional aspects.