Cryptocurrency futures forecasting and dynamic hedging: evidence from bitcoin and ether using time-varying volatility models

This study investigates the forecasting performance and risk management implications of competing volatility models applied to Bitcoin and Ether futures traded on the Chicago Mercantile Exchange (CME). Using daily data covering Bitcoin futures from December 2017 and Ether futures from February 2021 through April 2025, we estimate and compare Geometric Brownian Motion (GBM) specifications—including time-varying and jump-diffusion extensions—against Mixed-GARCH frameworks across the three closest maturity contracts. Model evaluation employs the Model Confidence Set (MCS) methodology with Mean Squared Logarithmic Error (MSLE) as the loss criterion. Our findings consistently identify ARMA(1,1)–GARCH(1,1) as the superior forecasting model across all contracts and horizons, while additional complexity from jump components or mean–variance feedback yields no meaningful improvement. Rolling conditional hedge ratios derived from the preferred model reveal systematic differences between the two markets: Ether futures exhibit higher rolling conditional hedge ratios and greater variance reduction relative to Bitcoin futures, implying tighter spot-futures co-movement and more effective risk transfer for Ether positions.