The article explores how different models can predict extreme market risk for daily returns on the DowJones index. They compared GARCH models and Extreme Value Theory (EVT) to forecast Value-at-Risk at 95% and 99% confidence levels. GARCH models performed well, especially with skewed-t distribution. EVT techniques outperformed GARCH, with unconditional EVT being the most accurate. Models using fat-tailed distribution passed coverage tests at both confidence levels, showing consistent performance over time.