Unlike a linear risk, the exposures of options are nonlinear, because they respond non-constantly to changes in the value of the underlying instrument depending on whether they are the money (ITM), at-the-money (ATM), or out-of-the-money (OTM). As the commercial products are continuously renovated, new financial instruments are more complicated than ever. The first option (warrant) in Taiwan was issued in 1997. Since the authoritative organizations, such as BIS, recommended the Value-at-Risk (VaR) as a way to quantify marketing risks, VaR has recently become an important tool on market risk management. In this paper taking options data from Taiwan in our empirical study, we calculate the VaRs of options by using Delta, Delta-Gamma, Monte Carlo (MC) Simulation, and Extreme Value Theory(EVT) four methods, and the estimated VaR and predicting effectiveness of these models are compared. Our conclusions include: (1) Both skewness and kurtosis of option is significant higher than that of its underlying asset. (2) The performance of Delta method is good for deep ITM option, but not for the case of OTM option. (3) The evaluation results from Delta and Delta-Gamma are quite similar due to the near zero Gamma of options. (4) MC method has the lowest error efficiency and also the highest average range. (5) In the case of OTM or ATM option, EVT performs precisely, but not in the case of ITM option. (6) In the case of OTM option, as it approach to expiration date, the option values will be all near zero which induce to a very small estimated volatility (and also the VaR). In such case a false outlier (|real return|>estimated VaR) can occur easily. Hence, in calculating the VaR of option, the near expiration date returns should be removed from data, especially in the case of OTM.