Abstract: In this paper the author points out that G(y)=exp( - e-y), y0, the distribution of magnitude of earthquakes, suggested by B. Epstein is not the Type I distribution of the largest value, but is the Type Poisson-Bxponential compound distribution of extreme value because the Epstein's distribution has a saltus e- at y=0. The method suggested by B. Epstein is not feasible if no earthquake is observed in a year from all data.The precent author not only gives the general theory and method, but also proposes a new method which is similar to the computation of the height of typhoon wave in ocean. This method has advantage for the area where earthquakes are observed not very often.