Efficient Interior-Node Recognition

Summary

This chapter re-examines the implications of how to test and score the game-theoretical decidability of positions during the search. We focus on the efficient yet seamless integration of such interior-node recognition into modern chess programs. By means of a thorough discussion of its fundamental principles, we reveal various problems related to the practical application of interior-node recognition. Based thereon, we present an implementation framework for recognizers that solves all known problems and has proven its practical viability in our high-speed chess program DarkThought.

Among others we introduce the new concept of material signatures which allow for a quick and easy classification of chess positions into different categories of material distribution. By including material signatures in the internal position representation of the chess board, they can incrementally be updated during the execution of moves. This makes the computation of material signatures extremely cheap in practice.

The chapter is a modified reprint of our article “Efficient Interior-Node Recognition” as published in the ICCA Journal 21(3), pages 156–167, September 1998. The original article won the ICCA Journal Award 1999 for the best paper by a Ph.D. student published in the issues 21(2)-22(1).