We apply two methods, historical simulation method and Monte Carlo method, to construct the Value-at-Risk (VAR) model. The data is the weekly stock price and the selected companies are based on Taiwan 50-index. Four companies, UPEC, FPG, TSMC, and YCPS, are chosen from Taiwan 50-index for the experiment. Results show that there is a dramatic error occurred when the historical simulation method was used. This is because of the political event on March 19th, 2004. However, the results of backtesting for the Monte Carlo simulation method show that it is more robust than historical simulation method.