Skip to content
All library documents

Deep Reinforcement Learning for Tax-Aware Stock Trading

Article arXiv papers · Author: Shan Huang
Read the original

Summary

The paper studies how capital gains taxes and tax benefits from realized losses affect stock-trading decisions. It proposes using the average tax basis in a reinforcement-learning environment, allowing an agent to learn strategies with taxes included as well as strategies that ignore them. The central lesson is that tax treatment can change the preferred timing and selection of trades, so it can be important to model taxes directly when optimizing continuous stock trading.

The reported comparison finds that ignoring taxes can reduce average portfolio returns by more than 62%. The excerpt does not describe the data, market, tax rules, reward design, or experimental setup behind that result, so its size and generality cannot be assessed from the available text. The finding supports accounting for taxes in strategy design, but does not establish that the same effect will hold across jurisdictions, tax situations, or trading horizons.

Key ideas

  • The method uses average tax basis to represent tax consequences in stock-trading decisions.
  • Reinforcement learning is used to learn strategies both with and without tax considerations.
  • The reported results indicate that ignoring taxes can substantially reduce average portfolio returns.
  • The excerpt provides too little detail to judge how broadly the reported effect applies.

Tags

EquitiesMachine learningRisk management
Full text
# Taxable Stock Trading with Deep Reinforcement Learning


# Taxable Stock Trading with Deep Reinforcement Learning









In this paper, we propose stock trading based on the average tax basis. Recall that when selling stocks, capital gain should be taxed while capital loss can earn certain tax rebate. We learn the optimal trading strategies with and without considering taxes by reinforcement learning. The result shows that tax ignorance could induce more than 62% loss on the average portfolio returns, implying that taxes should be embedded in the environment of continuous stock trading on AI platforms.

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.