In the following paper we describe a method for developing computer programs for endgames in chess. Rules based on knowledge about chess are their most important part. But α-β-search and storing positions together with their evaluation are also part of the programs. The programmer himself can decide about the influence of each of the three components: knowledge, search and table look-up. For instance, he may reduce the amount of computing time by using more rules, or on the other hand he may only use simple rules and burden the tree- search with almost all the work.
Additionally this paper shows how to validate these programs, i.e., how to prove their correctness. We illustrate the method by applying it to the endgame “King and Pawn against King” and “King and Queen against King and Queen”, but it can cope with others, too.
Our method does not require big data bases. Especially it is not necessary to generate a data base that contains all positions of the treated endgame and their evaluations.