Recent developments in nonequilibrium statistical mechanics suggest that the history of entropy production in a system determines the relative likelihood of competing processes. This presents the possibility of interpreting and predicting the self-organization of complex active systems, but existing theories rely on quantities that are challenging to obtain. Here, we address this issue for a general class of Markovian systems in which two types of self-replicating molecular assemblies (self-replicators) compete for a pool of limiting resource molecules within a nonequilibrium steady state. We derive exact relations that show that the relative fitness of these species depends on a path function, ψ, which is a sum of the entropy production and a relative-entropy term. In the limit of infinite path length, ψ reduces to the entropy production. We demonstrate use of the theory by numerically studying two models inspired by biological systems, including a simplified model of a competition between strains of the yeast prion Sup35 in the presence of driven disaggregation by the ATPase Hsp104.