Abstract: The predictive value of deep-learning models in finance depends critically on the temporal context supplied to the network, yet the literature lacks a systematic, multi-market evaluation of ...