The evaluation of the success level of an earthquake prediction method should not be based on approaches that apply generalized strict statistical laws and avoid the specific nature of the earthquake phenomenon. Fault rupture processes cannot be compared to gambling processes. The outcome of the present note is that even an ideal earthquake prediction method is still shown to be a matter of a ''chancy'' association between precursors and earthquakes if we apply the same procedure proposed by Mulargia and Gasperini [1992> in evaluating VAN earthquake predictions. Each individual VAN prediction has to be evaluated separately, taking always into account the specific circumstances and information available. The success level of epicenter prediction should depend on the earthquake magnitude, and magnitude and time predictions may depend on earthquake clustering and the tectonic regime respectively.