This paper examines copulas that best fits the equity returns. Using nine years data of daily returns of 30 representative stocks, this study finds that t copula unanimously dominates the goodness of fit criteria. The conclusion reveals the inappropriateness of using high-dimensional multivariate Gaussian distribution to model the dependence of asset returns, because the nested distribution underestimates the volatility and anomaly of asset performance. Furthermore, Gumbel, Clayton, and Frank copula do not capture the extreme value dependence among assets. The results suggest that the optimal procedure for Monte Carlo simulation of asset return is to fit the individual asset return marginal and model the dependence of asset trends through the copula.