‘Acting against the odds’ is acting against one’s evidence that one will succeed. Most of the discussion on the rationality of acting against the odds has focused on whether believing against the odds is rational. In this paper, rather, we address the question of whether acting against the odds can be rational and, if so, under what conditions. First, we provide the first axiomatic characterisation of the preferences of an individual who acts against the odds. Second, we analyse the normative implications of having preferences so characterised. We understand an individual who acts against the odds as an individual who prefers having the agency of increasing the likelihood of attaining a good outcome, which would otherwise be low. To exemplify, an individual might choose to pursue a PhD because she prefers increasing the chances of attaining an academic job by exerting effort. She would not experience such agency while attempting to be successful in other, less risky, careers. Having this preference, we show, is inconsistent with other theories of behaviour under risk such as, prominently, expected utility theory. We thus analyse what behavioural properties a person who acts against the odds satisfy and we specify in which cases this behaviour counts as rational, and in which cases it does not.