A model that couples large-scale plate motion loading with the dominant mechanical effects of fluids within faults shows that the system always evolves to a high degree of disorder in the fault zone stress state. This disorder comes about because of the additional degree of freedom in stress space allowed by including coseismic changes in the fault zone pore pressure. For a given set of initial conditions the long-term fault-averaged stress state equilibrates, resulting in effective properties of the system while maintaining continuous complexity. Stress state disorder organizes in physical space into spatially correlated regions of incipient failure. If the spatial extent of incipient failure is limited, small events occur and the system continues forward. Aperiodically, incipient failure covers a large portion of the fault, and failure of a single cell cascades into very large model earthquakes. A switch from static to dynamic friction at the onset of slip results in a slip-weakening model from stress redistribution during a (quasi-static) propagating elastic dislocation. The complexity inherent in the model is quantified with simple relationships for the average stress drop, average slip, and seismic moment of large events. The characteristic length relating average slip to stress drop is shown to be asymptotic and a function of the aspect ratio of the rupture and the degree of fault overpressure. The model shows that average slip increases with rupture length but at a reduced rate with ongoing slip, showing a smooth transition from L model to W model mechanics for very large events. Model catalogs are compared with the historical catalog of strike-slip events <Wells and Coppersmith, 1994>, and the good agreement between the model and observations of seismic moment and fault area shows that the model successfully predicts the average slip over rupture areas up to about 2500 km2.