We suggest a theoretical basis for the comparative evaluation of forecasts. Instead of the general assumption that the data is generated from a stochastic model, we classify three stages of prediction experiments: pure non-stochastic prediction of given data, stochastic prediction of given data,...

In this study I apply the recent advances in volatility estimation and forecasting to the price series of the FTSE 100, retrieved from the TRTH database. I use the recently introduced threshold bipower variation and the realized semivariance together with the usual realized variance and the...