Using chunking to solve chess pawn endgames
Artificial Intelligence
Information content of chess positions
ACM SIGART Bulletin
A methodology for learning players| styles from game records
International Journal of Artificial Intelligence and Soft Computing
| Hi-index | 0.00 |
This paper is an attempt to classify that which is important in chess. Other authors have chosen to make classifications relating to, for example, the material on the board, or whether there exists an attack on the king. We find that this is sufficient for an introductory treatment of the subject of advantages. However, the chess specialist understands that two apparently similar positions may, in fact, be quite different since there may be a very important difference in the features of each.Earlier attempts to develop test suites for evaluating human and computer chess strength [12, 8], although valuable, have had clear drawbacks in terms of their depth, range and number of positions examined. We present a taxonomy of positions in chess that require special knowledge. The taxonomy is what drives our selection of positions, and not vice-versa. It is easy to understand what a passed pawn is, and a bit of classic advice such as “Passed pawns must be pushed” makes use of the simple metric that a passed pawn becomes more valuable with each advance. However, there are outside passed pawns, protected passed pawns, blockaded passed pawns, and passed pawn masses. Each requires its own understanding, and frequently other features of a position can cause great variation in what may at first appear to be positions that should be treated very similarly.