The article compares different models for predicting financial market risk in different volatility states. The researchers tested six indexes from various countries using popular models like GARCH and EVT. They found that GARCH(1,1) with a standardized Student t distribution was the most reliable model overall. Nonparametric techniques performed well in low volatility but not in turbulent times. They also suggested using an extreme distribution threshold for EVT models and combining the best models for more accurate predictions.