Surprisingly few empirical studies address how computational thinking works "in the wild" or how games and simulations can support developing computational thinking skills. In this article, the authors report results from a study of computational thinking (CT) as evinced through player discussions around the collaborative board game "Pandemic" (Leacock, 2007). To better understand how CT practices may be embedded in certain forms of gameplay, the authors investigated "Pandemic" as a test-bed to differentially assess the relationship of computational thinking to minor modifications of the game. Through both quantitative coding of in-game talk and qualitative analysis of discourse exchanges, it was found that seemingly minor modifications to the game resulted in significant differences in the character and amount of computational thinking shown in player discourse. In particular, the data suggest that a "ghost player" modification--in which players share decision-making for one additional collective player in addition to their own roles--allowed students to voice both more logic and more complex logic. The analyses suggest that this improvement is due, in part, to a flexible relationship of identity to logic and action allowed by the ghost player modification, implying that the social components of board games can have provocative interactions with their computational systems.