Axiomatization of a Preference for Most Probable Winner