The model of conditional value-at-risk (CVaR) was utilized to control market risks.Wavelets technique and extreme value theory (EVT) were combined to estimate the conditional value-at-risk.Wavelets were used as a threshold in generalized Pareto distribution,and EVT was applied with a wavelet-based threshold,then the CVaR was estimated by virtue of the extreme theory.This new model has been applied to two major stock markets:the Hang Seng index and the Shenzhen composite index.The relative performance of the wavelet-based EVT was benchmarked against the conditional extreme value theory.The empirical results show that the wavelet-based EVT improves the predictive performance of financial forecasting according to the number of violations and for the results of tail-loss tests.