ARMA processes model financial return series by tracking a weighted moving average of randomized noise, together with a weighted moving average of past model outputs. These are "pretty good," in the sense that they exhibit decaying autocorrelation and covariance stationarity, but have constant variance. ARMA-GARCH processes improve on the model by replacing ARMA-level noise with a GARCH process, capturing time-dependent volatility and volatility clustering. This applet provides VaR and ES forecasts based on ARMA-GARCH modeling and their predicted conditional means and volatilities.