Matthew Saffell
Matthew Saffell is a machine-learning researcher who, with John Moody, co-authored the foundational reinforcement-learning trading literature produced at the Computational Finance Program of the Oregon Graduate Institute of Science and Technology. He is a co-author of the Journal of Forecasting paper Moody Wu Liao Saffell 1998 (with Moody, Lizhong Wu and Yuansong Liao), of the NIPS 1998 conference paper “Reinforcement Learning for Trading”, and of the definitive journal article Moody and Saffell 2001, “Learning to Trade via Direct Reinforcement” (IEEE Transactions on Neural Networks).
He appears in this vault as a co-originator of Recurrent Reinforcement Learning Trading and the direct-reinforcement approach to trading-policy optimisation — training a recurrent policy by gradient ascent to maximise a risk-adjusted performance function such as the Differential Sharpe Ratio net of transaction costs, rather than first forecasting prices. That body of work is the root of the modern Reinforcement Learning Trading Policy.
Matthew Saffell [defines] Recurrent Reinforcement Learning Trading
As with his co-author, the vault records the foundational/definitional status as confirmed while flagging that the profitability claims in the 1998-2001 papers are dated, single-group, un-replicated backtests on data ending in 1994/1996.
Connections
- Moody and Saffell 2001 — proposes_model, 2001, source: https://ieeexplore.ieee.org/document/935097/
- Moody Wu Liao Saffell 1998 — proposes_model, 1998, source: https://link.springer.com/chapter/10.1007/978-1-4615-5625-1_10
- John Moody — relates (long-running co-author), source: https://pubmed.ncbi.nlm.nih.gov/18249919/
- Recurrent Reinforcement Learning Trading — proposes_model, source: https://ieeexplore.ieee.org/document/935097/
- Differential Sharpe Ratio — proposes_model, source: https://proceedings.neurips.cc/paper_files/paper/1998/hash/4e6cd95227cb0c280e99a195be5f6615-Abstract.html
- Reinforcement Learning Trading Policy — proposes_model, source: https://ieeexplore.ieee.org/document/935097/