Preference-based systems (p.b.s.) describe interactions between nodes of a system that can rank their neighbors. Previous work has shown that p.b.s. converge to a unique locally stable matching if an acyclicity property is verified. In the following we analyze acyclic p.b.s. with respect to the self-stabilization theory. We prove that the round complexity is bounded by n/2 for the adversarial daemon. The step complexity is equivalent to (n^2)/4 for the round robin daemon, and exponential for the general adversarial daemon.