A weighted sum of return, downside risk, differential return, and the Treynor ratio is proposed as an RL trading reward, with simple gradient properties and anecdotal backtest results.
Moody and Matthew Saffell
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
A Risk-Aware Reinforcement Learning Reward for Financial Trading
A weighted sum of return, downside risk, differential return, and the Treynor ratio is proposed as an RL trading reward, with simple gradient properties and anecdotal backtest results.