How Computers Play Chess
David Levy, Monty Newborn · ICGA Journal · 1990
In both cases (depth 4, breadth 40 and depth 7, breadth 10) PSSS* performed superior to "Parallel staged SSS", as proposed by Campbell (1981).In comparison to PVS PSSS* behaved better for depth 4 and worse for depth 7.From the viewpoint of computer-chess practice it is regrettable that Kraas did not implement PSSS* in a tournament computer-chess program.So we can not judge how well it would compete with other chess tree-search algorithms.However, Hans-Joachim Kraas is aware of this point.In the last paragraph of his thesis he states self-critically (roughly translated):"All these considerations let me expect that the speedup of PSSS* will decrease (considerably) if instead of 'theoretical game trees' real chess trees, as generated by chess computers, are to be analysed.Investigations for the comparable PVS-algorithm underline this assessment: Only for small numbers of processors PVS achieves a satisfactory speedup.It will cost further efforts -for PVS as well as for PSSS* -to improve the performance of parallel systems in the analysis of chess trees."The thesis of Kraas is well organized and easy to read (in German).The ideas and also the technical parts are explained in detail.Previous work of other scientists is mentioned adequately and -as far as I can judge thiswithout gaps.I have profited from reading the thesis.