Ranking the Value at Risk (VaR) models to reduce the opportunity cost of capital requirements is as important tool for evaluating the performance of these models. In this paper we used a new method to rank and test the predictive ability of models. This method allows the risk managers to interpret effectively risk and cost of capital allocation by considering the size of the punishment measure for VaR models. We applied this method for comparing the performance of existing parametric VaR models such as GARCH Models and Markov regime switching, for daily data of Tehran stock exchange index in period 2001-2012. The results confirm that FIEGARCH and EGARCH models have better performance than the GARCH and Markov model in predicting VaR for Tehran stock exchange index.