Optimal Strategy in Werewolf Game: A Game Theoretic Perspective

Abstract

Werewolf game, also known as Mafia game, is a social deduction game that models the conflict between an informed minority (werewolf group) and an uninformed majority (citizen group). This paper explores the optimal strategies of the werewolf game from the perspective of game theory, focusing on cases both with and without prophet. First we examine the existing strategy in game without prophet and propose ``random strategy +", which provides an improved winning probability for the werewolve group. Then we further study the game with prophet, and find the game with prophet can be transformed into a extensive game with complete but imperfect information under a specific rule. We construct a model and design an algorithm to achieve PBE and maximize the citizen group's winning probability. In the end, we examine a property of PBE in game without any restriction.