As early as St Augustine and Aquinas, philosophers have studied the epistemology of contradictions to natural laws (aka, miracles). In this article, we develop a game-theoretic framework to analyze the concept of a miracle as understood by empiricists such as David Hume and John Locke. A proper distinction between natural laws and miracles cannot be drawn absent assumptions about the deity’s preference. We demonstrate that a deity aiming to maximize a Bayesian believer’s faith, would jointly produce uniform scientific evidence and miraculous subjective experiences. Higher confidence in scientific knowledge enhances, rather than undermines, religious belief because it facilitates an interpretation of miracles as supernatural intervention rather than natural events. We also show that a benevolent deity would favor persuasion by means of painful experience only for believers experiencing doubt, i.e., for intermediate levels of faith.