Automatic generation of opening books for dark chess

Abstract

Playing the opening game of dark chess well is a challenge that depends to a large extent on probability. There are no known studies or published results for opening games, although automatic generation of opening books for many games is a popular research topic. Some researchers collect masters' games to compile an opening book; while others automatically collect computer-played games as their opening books. However, it is difficult to obtain a strong opening book via the above strategies because few games played by masters have been recorded. In this paper, we propose a policy-oriented search strategy to build automatically a selective opening book that is helpful in practical game playing. The constructed book provides positive feedback for computer programs that play dark chess. © 2014 Springer International Publishing Switzerland.