Clues About Bluffing in Clue: Is Conventional Wisdom Wise?

We have used the board game Clue as a pedagogical tool in our course on artificial intelligence to teach formal logic through the development of logic-based computational game-playing agents. The development of game-playing agents allows us to experimentally test many game-play strategies and we have encountered some surprising results that refine “conventional wisdom” for playing Clue. In this article, we consider the effect of the oft-used strategy wherein a player uses her/his own cards when making suggestions (i.e., “bluffing”) early in the game to mislead other players or to focus on acquiring a particular kind of knowledge. We begin with an intuitive argument against this strategy together with a quantitative probabilistic analysis of this strategy's cost to a player; both suggest “bluffing” should be detrimental to winning the game. We then present our counterintuitive simulation results from playing computational agents that “bluff” against those that do not that show “bluffing” to be beneficial. We conclude with a nuanced assessment of the cost and benefit of “bluffing” in Clue that shows the strategy, when used correctly, to be beneficial, and, when used incorrectly, to be detrimental.