A common feature of many speculative attack models on currencies is the existence of multiple equilibrium solutions. When choosing the equilibrium strategy, a trader faces Knightian uncertainty about the rational choice of the other traders. We show that the concept of Choquet expected utility maximization under Knightian uncertainty leads to unique equilibria. In games of incomplete information the optimal strategy maximizes the expected utility with respect to a two-dimensional information: environment and rationality. We define a new concept of equilibria, the Choquet-expected-Nash-equilibria, which allows the analysis of decisions under uncertainty, which result in multiple equilibria in standard analysis. We provide uniqueness theorems for a wide class of incomplete information games including global games and apply them to fairly general currency attack models. The uniqueness of the equilibrium remains valid for arbitrary noise distributions, positively correlated signals, the existence of large traders, individual payoff functions, and for the case that non attacking traders suffer a loss in case of a successful attack, as is the case for investors in the attacked country.