The criterion for assessing the accuracy of coal reserve estimates proposed by Philip and Watson (1982) is criticized.Following Royle (1977), it is suggested that the global estimation variance found by geostatistical methods is more appropriate.This method is applied to coal data from the Bowen Basin. This shows that two factors (the inherent variability of the variable under study and the total number of drill holes) have more influence over the accuracy of the reserve estimate than the form of the drilling pattern. Comparisons between the relative errors for different sized regions demonstrate that the spacing between drill holes is of secondary importance. The dominant factor is the total number of holes.Comparisons between the relative errors show that some variables such as sulphur need a much denser sampling than the others to obtain the same level of precision.