Abstract: In view of the irrationality of the normal hypothesis of the value-at-risk model， the POT model based on the generalized pareto distribution is used to measure the tail risk of China's stock market， and the influence of stock liquidity on the tail risk of China's stock market is explored by grouping the sample data according to liquidity. In the research， the slope change point theory in statistical theory is introduced into the traditional threshold selection method， which effectively avoids the subjectivity problem when determining the optimal threshold. The test results show that for the measurement of tail risk of the stock market， the VaR value at the 95% confidence level is more reliable than the estimated result of the VaR value at the 90% confidence level， and the estimated value shows that tail risk and liquidity change in the opposite direction， that is， the smaller the stock liquidity is， the greater the market tail risk.

Key words: stock liquidity; tail risk; ARMA-GARCH model; POT model