An arbitrary dependence structure between a finite family of events of a probability space defines a hypergraph structure. We study the converse operation, starting from a hypergraph structure, to determine a canonical probability space with events indexed by the vertices and that satisfy the prescribed dependence structure; we also require that events that are not mutually independent must be mutually exclusive. Such a hypergraph structure, we call a configuration. Requiring furthermore that all events are given the same positive probability $t$, converts this problem to the study of a family of Möbius polynomials, which we call the relative Möbius polynomials (or independence polynomial) of the configuration. This study was very much already complete in the case where the configuration actually derives from a graph structure, but remained open in the general case. We show the existence and uniqueness of an optimal real $t$, optimal in the sense that the events cover a subset of maximal probability. We base our study on a new formula for the derivative of the Möbius polynomial.