We present a volatility forecasting comparative study within the ARCH class of models. Our goal is to identify successful predictive models over multiple horizons and to investigate how predictive ability is influenced by choices for estimation window length, innovation distribution, and...