In recent years, refinement of stochastic storm surge estimation is essential for risk management in insurance industries because the Japanese government promotes flood risk transfer to insurance companies. Insurance systems may reach peak risk when storm surge damage occurs; however, there are only a few studies on the stochastic analysis of storm surges. This paper presents the stochastic evaluation of storm surges in Tokyo Bay. First, storm surges are assessed using two methods of an empirical formula and a numerical model. Then, the return period of storm surges is stochastically evaluated. It is found that an empirical formula underestimates the surge level in comparison to the numerical model. Based on the results of numerical model, the return period of a storm surge is proposed in Tokyo Bay.