Geodetic and geologic observations from a handful of well-studied great plate boundary earthquakes provide a basis for exploring those features of the deformation cycle having the strongest influence on recurrence estimation. In ideal circumstances the time-predictable model requires knowledge of only the seismic slip (and time) of the last earthquake and the rate of relative plate motion or fault slip. Practically, indirect measurements must usually suffice, local coseismic strain changes rather than seismic slip and local deformation rate rather than slip rate. Permanent, nonrecoverable deformation, often roughly half the coseismic strain offset near great thrust earthquakes, complicates the inference of interplate seismic slip. Owing to short- and long-term postseismic transients, commonly 20--40% of the coseismic strain drop, deformation rates are variable and not simply related to the plate motion rate. At strike-slip plate boundaries these complications can sometimes be avoided and seismic slip and slip rate can be obtained more or less directly. When these favorable circumstances are absent and when knowledge of the deformation cycle is incomplete, recurrence accuracy can be improved by empirically correcting for unknown elements of the cycle. Systematic features in the spatial distribution of the transients and permanent deformation near subduction zones help identify the regions where these corrections are largest and where the shortcoming of the empirical approach are most severe.