We have constructed a synthetic seismicity model of the Wellington region, New Zealand, including seven superfaults and 500 subfaults which are randomly positioned. From this model, a synthetic catalogue of 2000 years duration, containing events of magnitude 5.0 or more, has been generated. The properties of the catalogue, such as the long-term slip rates, b value, average activity rate, and hypocenter distribution, are in accord with paleoseismic studies and the real seismicity over the last 40 years. Such a synthetic catalogue can replace the short, incomplete, and inhomogeneous historic and instrumental records in research which needs a long time duration and many strong shocks. We have used our catalogue to examine tests for the existence of accelerating moment release (AMR) before large events and compared the results with those from random (Poisson) catalogues. We find that (1) the apparent success rate is very dependent on the rules used to define the test window; (2) when appropriately defined, the AMR pattern occurs before about 20% of the strong (M ≥ 7.0) shocks with a typical precursor time of about 22 years; (3) the AMR pattern is found almost equally frequently before large events in random catalogues; and (4) there are some false alarms (AMR pattern without a large event). This extended study reinforces the conclusion in our preliminary report: that in synthetic catalogues of the kind we have constructed, the AMR pattern is essentially an artefact of the method of sampling.