This paper clarifies the relation between case-based prediction and analogical transfer. Case-based prediction consists in predicting the outcome associated with a new case directly from its comparison with a set of cases retrieved from a case base, by relying solely on a structured memory and some similarity measures. Analogical transfer is a cognitive process that allows to derive some new information about a target situation by applying a plausible inference principle, according to which if two situations are similar with respect to some criteria, then it is plausible that they are also similar with respect to other criteria. Case-based prediction algorithms are known to apply analogical transfer to make predictions, but the existing approaches are diverse, and developing a unified theory of case-based prediction remains a challenge. In this paper, we show that a common principle underlying case-based prediction methods is that they interpret the plausible inference as a transfer of similarity knowledge from a situation space to an outcome space. Among all potential outcomes, the predicted outcome is the one that optimizes this transfer, i.e., that makes the similarities in the outcome space most compatible with the observed similarities in the situation space. Based on this observation, a systematic analysis of the different theories of case-based prediction is presented, where the approaches are distinguished according to the type of knowledge used to measure the compatibility