The researchers used a method called bivariate extreme value theory to study the extreme behavior of the Shanghai and Shenzhen stock markets. They developed a new extreme copula called t-EV-copula, which accurately captures extreme data and both upper and lower dependencies between the markets. By using this copula, they were able to model the tail joint distribution of the markets and visualize it. Finally, they used VaR as a measure to describe the tail risk of the joint distribution.