The article examines different models to estimate market risk for the WIG20 index. By analyzing return rates from 1999-2011, the researchers used ARIMA-GARCH models to calculate Value at Risk (VaR) for both long and short positions. They tested various conditional distributions (normal, t-student, GED) to find the best model for VaR estimation. The Kupiec test was used to compare the accuracy of the models.