This dissertation examines why powerful states, acting as third parties in international disputes, succeed or fail in efforts to end war. I develop a game-theoretic model that shows how bias a close alignment of preferences between a major power and one of the disputants—affects the manner and the outcome of intervention bargaining. The formal model suggests that biased powers select coercion as a method of intervention, while impartial powers choose to mediate. In addition, the model shows that bias reveals private information about an intervener's willingness to secure an agreement using force. When a highly biased power intervenes in a crisis, a peaceful settlement is likely because disputants are certain the intervener will enforce an agreement by military means. When a third party shows less favoritism, negotiations fail because disputants doubt such a party is committed to use force. Peace is again more likely when an intervener is impartial because it behaves as a mediator, seeking agreements both adversaries find acceptable. To test the hypotheses I derive from my formal model, I conduct a statistical analysis of 290 crises since 1945 and a qualitative inquiry into U.S. and British interventions in three Balkan crises: the Italo-Yugoslav dispute over the port city of Trieste (1945), the final Trieste settlement (1954), and the Serbo-Albanian conflict over the Kosovo province (1995-1999). As my theory predicts, the empirical evidence indicates that a peace settlement is most likely when the powerful third party favors both disputants or when it is significantly biased against one, but for different reasons. A great power that favors both actors is very motivated to find a solution that satisfies both sides, whereas a highly biased power has a credible threat to use force against the less favored belligerent.