Deep Learning for Financial Time Series: A Large-Scale Benchmark of Risk-Adjusted Performance

Abstract

We present a large-scale benchmark of modern deep learning architectures for financial time-series prediction and position sizing, with a primary focus on Sharpe-ratio optimization. Evaluating linear models, recurrent networks, transformer-based architectures, state-space models, and recent sequence-representation approaches, we assess out-of-sample performance on a daily futures dataset spanning commodities, equity indices, bonds, and FX from 2010 to 2025. Beyond average returns, we evaluate statistical significance, downside and tail-risk measures, breakeven transaction-cost analysis, robustness to random seed selection, and computational efficiency. We find that models designed to learn rich temporal representations consistently outperform linear baselines and generic deep learning models. Hybrid variants such as VSN-LSTM attain the strongest overall Sharpe ratio, while VSN-xLSTM and LSTM-PatchTST combinations show strong downside-adjusted characteristics.

Kieran Wood

DPhil in Machine Learning, Oxford · AI Solutions Architect, NVIDIA

Latent regime structure and non-stationary time series; deep learning for forecasting and sequential decision-making.