The article examines different methods for estimating the risk of investment portfolios using stocks, commodities, and FX futures. It looks at which mathematical models work best for predicting potential losses in these portfolios. The study finds that models using GARCH-margins are more accurate than traditional correlation-based models for estimating Value at Risk (VaR). However, it also shows that no single model is consistently the best for all situations. In fact, most models underestimated the actual risk of the portfolios. The research suggests that finding the best model depends on the type of risk being measured and can change over time.