Abstract
In the context of a three-part study of the KQKR endgame, some heuristics appealing to human players are examined. Three measures of goodness are defined, by means of which these heuristics can be evaluated in a uniform way, given complete knowledge of the domain. The evaluation proves that, as judged by these objective measures, apparently intuitively attractive heuristics vary greatly in their utility. The study of these heuristics and some further observations are shown to be a reasonable basis for at least a crude predictor of human play. The principles and limitations of such a programmed predictor of human counterplay are discussed.
Get full access to this article
View all access options for this article.