The Rock-Scissors-Paper game strategy estimation method using homology search Takahiro Komai 1 , Song-Ju Kim 2 , Takuji Kousaka 3 , and Hiroaki Kurokawa 1 1 School of Computer Science, Tokyo University of Technology, 1404-1 Katakuramachi Hachioji-City Tokyo JAPAN 2 Graduate School of Media and Governance, Keio University, 5322 Endo Fujisawa Kanagawa 252-0882 JAPAN 3 Electrical and Electronic Engineering, Chukyo University, 101-2 Yagoto Honmachi, Showa-ku, Nagoya-shi, 466- 8666, Japan Abstract The Purpose of this study is to propose a method to estimate Rock-Scissors-Paper (RSP, Janken) game strategy based on human personality in RSP game. In order to estimate the player's strategy in the RSP game, it is considered effective to compare the player's hand sign series and the hand sign series obtained from various RSP strategies based on simi- larity. Therefore, in this study, focused on homology search as a method of calculating similarity between sequences. This paper shows the result of verifying the analysis accu- racy of homology search against RSP strategy. 1. Introduction In recent years, human behavior prediction has been studied in various fields. In human behavioral research, many estimation methods using that person's past behav- ioral data have been proposed [1][2]. In the behavior of a person, there is a decision-making rule that determines the behavior of the person, i.e. "behav- ior strategy". And each person has a behavioral strategy based on personality to solve the given problem. It is thought that it is possible to predict the behavior of the per- son if it can estimate the person's behavioral strategy. How- ever, because each person has different personality, human behavior prediction is a difficult problem. Therefore, in this research, we focused on Rock-Scis- sors-Paper (RSP) geme as a problem that behavioral strat- egy strongly depends on person's personality. The RSP game is one of the games that competes for its victory and defeat by using three kinds of hand signs (Rock, Scissors and Paper) in the three-cornered deadlock relationship. In this research, we use the RSP game which repeats the con- stant times game and competes for the total number of vic- tories. In previous study, we considered the RSP strategy to de- termine the player's hand sign as a set of simple strategies. Then, the following four strategy algorithms were assumed as the strategy that becomes the element. 1. Simple pattern strategy : SP strategy 2. SP strategy with random elements : SP_R strategy 3. Probability strategy : Pb strategy 4. Probability transition strategy : Pb_T strategy We have proposed a method for predicting the future hand sign from past hand sign series using the hand sign series obtained from these strategy algorithms. As a result of the experiment, it is known that a method using genetic pro- gramming is effective for strategies 1 and 2, and a method using the tug-of-war dynamics is effective for strategies 3 and 4 [3]. Effective behavior prediction methods differ according to opponent strategy as described above. From this result, if it is possible to estimate the strategy of the opponent, more effective behavior prediction method can be applied. Then, the accuracy of behavior prediction can be improved. Therefore, the purpose of this research is to propose a method to estimate the strategy of the player in the RSP game. In order to estimate the player's RSP strategy, it is con- sidered that effective to compare player's hand sign series and hand sign series obtained from each strategy algorithm based on similarity. Therefore, in this research, we focused on the method for calculating the similarity between se- quences used in homology search. First, we explain the four strategy algorithms assumed as elements of human's RSP strategy. Next, conventional re- search is described. Finally, we explain the RSP strategy estimation method proposed in this research. And, it shows the results of verifying whether the proposed method is ef- fective for estimating the RSP strategy. 2. Human strategy in RSP game Here, we explain four strategy algorithms assumed as el- ements of human strategy in RSP game. 2.1. Simple Pattern (SP) strategy The SP strategy is a strategy that repeats outputting of a short patterns of Rock, Scissors, Paper which is hand sign 2019 RISP International Workshop on Nonlinear Circuits, Communications and Signal Processing (NCSP2019) Honolulu, Hawaii, USA, March 4-7, 2019 068