The article deals with Value-at-Risk (VaR), by using method, which does not require defining distribution of return of the financial instrument. We used Cornish- Fisher expansion, which allows evaluating the quintiles and VaR of the explored return distribution only by knowing its characteristics of skewness and kurtosis. This method allows achieving sufficient accuracy of VaR even from relatively small amount of experimental data of return. The empirical evaluation of VaR, by applying Cornish-Fisher expansion, is performed by using hourly trading data of EUR/USD long position.