In recent years, several oligopolistic models of the liberalized power exchanges have been proposed, within two paradigms based upon radically different assumptions on rationality, learning and cognition: optimal choice and agent-based computational modeling. This paper is a first attempt to compare the explanatory performances of an agent-based model with a supply function equilibrium model on the same dataset. The models are designed in such a way that differences in performance between them are mainly due to their different behavioral assumptions, and are validated on a unique plant-level dataset on the Italian power exchange. As suggested by our findings, the agent-based model is better able to capture the intraday profile of power prices, but both models tend to overestimate the degree of competition among generating companies.