Implementation and qualitative analysis of an adaptive computer Shogi program by producing seesaw game

This paper reports on a novel method for dynamic adjustment of the skill level of a Shogi program for each opponent player and an analysis of its effects on the game experience. Recently, there have been several studies on adaptive game artificial intelligence (AI). However, the existing methods cannot be applied to games that require deep and narrow search to calculate the best moves. In addition, the question of which main element motivates the players, remains unsettled. Therefore, we have developed a computer Shogi program that controls the game position as seesaw game progress. More than 6000 games against the proposed system and almost 50 responses to a questionnaire were gathered. The results from the statistical and qualitative analysis indicated that the proposed system adjusted itself to each player in terms of the winning rate and in a subjective evaluation of strength. Furthermore, we concluded that the seesaw game progress motivated players without the need to adjust the winning rate to 50%.