Jo Nov 7, 2023

Game theory is a set of algebraic models to investigate behavior conflicting with each other. In practice, event players often have a game in uncertain conditions with lack of opponents’ strategies.

Especially, in such games as two-person zero-sum games, the payoff value of each player’s strategy is imperfect, so high accuracy of player modeling cannot be guaranteed. In general, statistical method processes uncertainty mathematically with lots of observed numerical value data, but observed data usually contain occasional fluctuation caused instantly by various reasons as well as essential information about opponents.

Being used as a technology of several variables which identifies common characteristic factors in an invariable set, statistical factor analysis uses probability methods for factor analysis. However, these methods include invariable assumptions about probability distribution so the results may not always be right. Also, there are some cases where modeling is not suitable in practice. Though modeling is not correct strictly, we can draw nearly right conclusions from those hypotheses if it is approximately suitable.

Game theoretical decision adoption is a decision-making problem which determines how the best result can be obtained when the result depends both on its own and on the opponent’s actions.

Some decision-making problems like a zero-sum game are inflexible in target weight according to the structure, so they are changed and affected by opponents. Information analysis methods used in the past are not sufficient to deal with these problems.

Kim Ok, a researcher at the Faculty of Information Science and Technology, has proposed an algorithm where a player’s model can be estimated by a small amount of game data in a two-person zero-sum game characterized by a monotonically decreasing function.

The algorithm is based on the analysis of game process profiles. First, she introduced ordered weight operators to the strategic sequence determination of players and defined a compensation function considering monotonically decreasing character of a two-person zero-sum game. Then, she evaluated the game payoff value of the strategy and conducted player modeling.

The proposed algorithm will contribute to enhancing the accuracy of player modeling in two-person zero-sum games.