A methodology is presented for computing elastic-rebound-based probabilities in an unsegmented fault or fault system, which involves computing along-fault averages of renewal-model parameters. The approach is less biased and more self-consistent than a logical extension of that applied most recently for multisegment ruptures in California. It also enables the application of magnitude-dependent aperiodicity values, which the previous approach does not. Monte Carlo simulations are used to analyze long-term system behavior, which is generally found to be consistent with that of physics-based earthquake simulators. Results cast doubt that recurrence-interval distributions at points on faults look anything like traditionally applied renewal models, a fact that should be considered when interpreting paleoseismic data. We avoid such assumptions by changing the "probability of what" question (from offset at a point to the occurrence of a rupture, assuming it is the next event to occur). The new methodology is simple, although not perfect in terms of recovering long-term rates in Monte Carlo simulations. It represents a reasonable, improved way to represent first-order elastic-rebound predictability, assuming it is there in the first place, and for a system that clearly exhibits other unmodeled complexities, such as aftershock triggering.