Skip to content
All library documents

Deep Reinforcement Learning for Trading Across Markets

Article arXiv papers · Author: Jędrzej Maskiewicz et al.
Read the original

Summary

The paper evaluates two deep reinforcement learning methods, Double Deep Q-Network and Proximal Policy Optimization, for trading. Their performance is compared with a buy-and-hold benchmark using daily data from 2019 to 2023. The assets described include three currency pairs, the S&P 500 index, and Bitcoin.

The reported results indicate that the reinforcement learning systems can avoid trades in unfavorable conditions and achieve better risk-adjusted returns than classical supervised-learning approaches. This suggests that a policy able to choose when to stay out of the market may contribute to risk management as well as trade selection. The summary does not specify the currency pairs, model design, transaction costs, validation procedure, or individual asset results, and it provides no numerical performance measures. The findings should therefore be read as claims about the study’s tested data and setup, not as evidence that the methods will generalize to other periods or live trading.

Key ideas

  • The study compares DDQN and PPO with a buy-and-hold benchmark.
  • It evaluates strategies on daily observations across currency pairs, an equity index, and Bitcoin.
  • The reported testing period runs from 2019 through 2023.
  • The authors attribute part of the methods’ risk management to avoiding unfavorable trades.
  • Reported risk-adjusted returns exceed those of classical supervised-learning approaches in the study.

Tags

Machine learningEquitiesForexCryptoRisk management
Full text
# Can Artificial Intelligence Trade the Stock Market?


# Can Artificial Intelligence Trade the Stock Market?









The paper explores the use of Deep Reinforcement Learning (DRL) in stock market trading, focusing on two algorithms: Double Deep Q-Network (DDQN) and Proximal Policy Optimization (PPO) and compares them with Buy and Hold benchmark. It evaluates these algorithms across three currency pairs, the S&P 500 index and Bitcoin, on the daily data in the period of 2019-2023. The results demonstrate DRL's effectiveness in trading and its ability to manage risk by strategically avoiding trades in unfavorable conditions, providing a substantial edge over classical approaches, based on supervised learning in terms of risk-adjusted returns.

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.