We present a new approach to obtaining empirically based (EB) criteria for estimating the epicentral location accuracy (i.e., ground truth, GT) of seismic events recorded at local distances on a regional network. The approach has been developed using a jackknife resampling method applied to carefully picked Pg phase arrival times for GT reference events from several South African gold mines. The events were well recorded locally by Southern African Seismic Experiment (SASE) stations within the Archean Kaapvaal craton, an area of relatively simple crustal structure. The region-specific criteria obtained specify an EBGT395% level of epicentral accuracy if events are recorded on eight or more stations at distances less than the Pg/Pn crossover (215 km) when the stations have a primary azimuthal gap <202 degrees. In addition, when nine or more stations are used for event location and one of them is within 79 km of the event, we find that a focal depth accuracy of 4 km at the 95% confidence level can be obtained and that an accuracy of 6 km can be obtained if eight stations are used. This result illustrates that GT criteria commonly applied to global event catalogs can be relaxed if an accurate velocity model and carefully picked phase-arrival times are used for event locations. Consequently, it is likely that additional events can be added to GT compilations by developing EBGT criteria for other regional networks and using them to identify candidate GT events. For example, the EBGT criteria developed in this study, when applied to the SASE seismicity catalog, yields 10 new GT events.