Jointly Learning Time-Series Momentum and Volatility for Futures Portfolios

Summary

This study applies deep multi-task learning to diversified time-series momentum portfolios in continuous futures. Its central idea is to learn portfolio construction and volatility-related tasks together, rather than treating the momentum signal and volatility estimate as independent inputs. Auxiliary tasks include forecasting realized volatility using different volatility estimators, with the aim of improving risk-adjusted portfolio decisions.

The approach is backtested from January 2000 through December 2020 and compared with existing time-series momentum strategies. The document reports that it outperforms those strategies after accounting for transaction costs of up to 3 basis points, and that adding auxiliary tasks improves portfolio performance. These are backtest findings on the stated futures portfolio and period; the summary does not describe the precise network design, asset universe, evaluation protocol, or robustness checks. Its results therefore do not establish that the method will perform similarly in other markets or live trading.

Key ideas

• The method jointly learns momentum portfolio construction and volatility-related auxiliary tasks.
• The study treats signal quality and volatility estimation as connected learning problems.
• Backtests cover continuous futures from January 2000 through December 2020.
• The reported results show outperformance over existing time-series momentum strategies after stated transaction costs.
• The supplied summary does not specify the model design or broader robustness evidence.

Tags

# Constructing Time-Series Momentum Portfolios with Deep Multi-Task Learning


# Constructing Time-Series Momentum Portfolios with Deep Multi-Task Learning









A diversified risk-adjusted time-series momentum (TSMOM) portfolio can deliver substantial abnormal returns and offer some degree of tail risk protection during extreme market events. The performance of existing TSMOM strategies, however, relies not only on the quality of the momentum signal but also on the efficacy of the volatility estimator. Yet many of the existing studies have always considered these two factors to be independent. Inspired by recent progress in Multi-Task Learning (MTL), we present a new approach using MTL in a deep neural network architecture that jointly learns portfolio construction and various auxiliary tasks related to volatility, such as forecasting realized volatility as measured by different volatility estimators. Through backtesting from January 2000 to December 2020 on a diversified portfolio of continuous futures contracts, we demonstrate that even after accounting for transaction costs of up to 3 basis points, our approach outperforms existing TSMOM strategies. Moreover, experiments confirm that adding auxiliary tasks indeed boosts the portfolio's performance. These findings demonstrate that MTL can be a powerful tool in finance.

Shown in full with attribution under the source's licence. Licence: abstract CC0

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.