Still crazy after all these years: the returns on carry trade

This paper proposes a novel approach to provide directional forecasts for carry trade strategies; this approach is based on Support VectorMachines (SVM), a Learning algorithm which delivers extremely promising results. Building on recent findings of the literature on carry trade we condition the SVM on indicators of uncertainty and risk; we show that this provides a dramatic improvement of the performance of the strategy, in particular during periods of financial distress such as the recent financial crises. Disentangling between measures of risk we show that the best performances are obtained by conditioning the SVM on measures of liquidity risk rather than on market volatility