Skip to content
All library documents

Testing Deep Reinforcement Learning Strategies for Crypto Backtest Overfitting

Article arXiv papers · Author: Berend Jelmer Dirk Gort et al.
Read the original

Summary

The paper treats backtest overfitting in deep reinforcement learning (DRL) for cryptocurrency trading as a hypothesis-testing problem. It trains DRL agents, estimates each agent’s probability of overfitting, and rejects agents judged to be overfitted. The aim is to reduce false positives from historical testing and improve the odds that selected strategies will perform well beyond the backtest.

Key ideas

  • The authors formulate detection of backtest overfitting as a hypothesis test.
  • They estimate overfitting probabilities for trained DRL agents and reject agents identified as overfitted.
  • In a test of 10 cryptocurrencies, less-overfitted agents had higher returns than more-overfitted agents and the stated benchmarks.
  • The evaluation covered a short period that included two crypto market crashes, so it does not establish performance across other market conditions.
  • Backtest screening offers evidence for agent selection but does not guarantee real-market success.

Tags

CryptoMachine learningStatisticsBacktestingRisk management
Full text
# Deep Reinforcement Learning for Cryptocurrency Trading: Practical Approach to Address Backtest Overfitting


# Deep Reinforcement Learning for Cryptocurrency Trading: Practical Approach to Address Backtest Overfitting









Designing profitable and reliable trading strategies is challenging in the highly volatile cryptocurrency market. Existing works applied deep reinforcement learning methods and optimistically reported increased profits in backtesting, which may suffer from the false positive issue due to overfitting. In this paper, we propose a practical approach to address backtest overfitting for cryptocurrency trading using deep reinforcement learning. First, we formulate the detection of backtest overfitting as a hypothesis test. Then, we train the DRL agents, estimate the probability of overfitting, and reject the overfitted agents, increasing the chance of good trading performance. Finally, on 10 cryptocurrencies over a testing period from 05/01/2022 to 06/27/2022 (during which the crypto market crashed two times), we show that the less overfitted deep reinforcement learning agents have a higher return than that of more overfitted agents, an equal weight strategy, and the S&P DBM Index (market benchmark), offering confidence in possible deployment to a real market.

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.