2019 Volume 36 Issue 1 Pages 119-151
IPSJ (Information Processing Society of Japan) conducts international artificial intelligence programming contests every year since 2012. I participated this contest in 2015 and 2016, and won the second prize in the 2015 contest and won the first prize in the 2016 contest. The subject of the 2015th and 2016th contests was a game with incomplete information. Basic rules of two contests were common, but features of the game were different. This paper introduces and evaluates four software methods of the 2016 champion software in comparison with those of the 2015 contest and related works. Four software methods are as follows: a method to construct a game tree for game with incomplete information by introducing concepts of meta node and meta action, a method for speedup bitmap computing by using SIMD (Single Instruction Multiple Data) and BMI (Bit Manipulation Instruction sets), a method to analyze incomplete information precisely, and a method to calculate evaluation value of the game by introducing concepts such as meta evaluation value. This paper also presents several findings gained by participating two contests.