I am doing an experiment using GARCH family models (GARCH, GJR-GARCH and GARCH) using normal and students t-distribution. I am predicting VaR for 252 trading days using a rolling window on data from 4 major stock indexes. the problem is that the R code I have used docent yield results that in theory should be possible. The packed I used us rugarch. I am now wondering if there is some one that know any other solutions in R or in python that can help me do this experiment.

What our results said was that students t-distrbution gave VaR values lower than normal which is not possible when the degree of freedom is around 5. thus there most be a problem with the packaged.