By using the last nonparametric estimation of VaR and ES, not depending on any distribution, analyze empirically the risk of Shanghai stocks index, explore the method of computing N-day VaR. The results of the exploration indicate that under more samples, there are better results by using nonparametric estimation of VaR and ES. the method of computing N-day VaR by curvilinear regression has better result than that of computing under the normal assumption.