Trading Signals in VIX Futures

We propose a new approach for trading VIX futures. assume that the term structure of futures follows Markov model. Our strategy selects position in by maximizing expected utility day-ahead horizon given current shape and level structure. Computationally, we model functional dependence between curve, positions, as deep neural network with five hidden layers. Out-of-sample backtests suggest this gives rise to reasonable portfolio performance, positions which investor will be either long or short contracts depending on market environment.