Paper: arXiv 2511.20678
Authors: Kamal Paykan
Abstract
This paper proposes a reinforcement learning–based framework for cryptocurrency portfolio management using the Soft Actor–Critic (SAC) and Deep Deterministic Policy Gradient (DDPG) algorithms. Traditional portfolio optimization methods often struggle to adapt to the highly volatile and nonlinear dynamics of cryptocurrency markets. To address this, we design an agent that learns continuous trading actions directly from historical market data through interaction with a simulated trading environment. The agent optimizes portfolio weights to maximize cumulative returns while minimizing downside risk and transaction costs. Experimental evaluations on multiple cryptocurrencies demonstrate that the SAC and DDPG agents outperform baseline strategies such as equal-weighted and mean–variance portfolios. The SAC algorithm, with its entropy-regularized objective, shows greater stability and robustness in noisy market conditions compared to DDPG. These results highlight the potential of deep reinforcement learning for adaptive and data-driven portfolio management in cryptocurrency markets.
Complexity vs Empirical Score
- Math Complexity: 7.5/10
- Empirical Rigor: 8.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced mathematical concepts like actor-critic architectures, entropy regularization, and stochastic control, while also demonstrating strong empirical rigor through backtesting on real cryptocurrency data from 2016-2024 with multiple risk-adjusted metrics.
Research Flowchart
flowchart TD A["Research Goal: Apply RL to Cryptocurrency Portfolio Management"] --> B["Methodology: Data Simulation & Environment"] B --> C["Input: Historical Market Data<br>(BTC, ETH, XRP, etc.)"] C --> D["Training Agents: SAC & DDPG Algorithms"] D --> E["Process: Learning Continuous Trading Actions<br>to Maximize Returns & Minimize Risk"] E --> F["Key Findings & Outcomes"] F --> G["SAC & DDPG Outperform Baselines<br>(Equal-weight, Mean-Variance)"] F --> H["SAC shows superior stability<br>via Entropy Regularization in volatile markets"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a reinforcement learning‑based framework for cryptocurrency portfolio management using the Soft Actor‑Critic (SAC) and Deep Deterministic Policy Gradient (DDPG) algorithms.
Source sentence
This paper proposes a reinforcement learning--based framework for cryptocurrency portfolio management using the Soft Actor--Critic (SAC) and Deep Deterministic Policy Gradient (DDPG) algorithms.
Assumes 4
- Classical financial models often assume normally distributed returns and stationary market conditions. statistical
Source sentence
Classical financial models often assume normally distributed returns and stationary market conditions—assumptions that are frequently violated in cryptocurrency markets.
- Daily portfolio rebalancing is possible without slippage or market impact. execution
Source sentence
The environment simulates daily portfolio re‑balancing, following the RL interaction loop.
- The reward can be defined as the log‑return of portfolio value. model
Source sentence
The simplest reward is log‑return: rt+1 = ln ( vt+1 vt ).
- The market data (OHLCV) from yfinance accurately reflects true price movements. data
Source sentence
Daily OHLCV data are obtained using yfinance.
Data
2016 – 2024 · BTC, ETH, LTC, DOGE · crypto · daily
- Four major cryptocurrencies (BTC, ETH, LTC, DOGE) OHLCV data market-prices public
Source sentence
Experiments use four major cryptocurrencies: BTC, ETH, LTC, and DOGE. Daily OHLCV data are obtained using yfinance. The period spans: Training: 2016‑2022, Testing: 2023‑2024.
Costs
Modeled
A constant cost rate c is applied to the sum of absolute changes in portfolio weights each period.
Source sentence
Accounting for transaction costs with rate c: Rnet t+1 = Rt+1 −c \nX i=1 |w(i) t −w(i) t−1|.
Finds 4
Sharpe 0.07 · total return 176.3%
- Both RL agents outperform the Markowitz MPT benchmark, with SAC delivering the highest risk‑adjusted performance. performance
Source sentence
Both reinforcement learning agents outperform the MPT benchmark, with SAC consistently delivering superior risk‑adjusted performance.
- SAC shows smoother convergence, lower tail risk (VaR -0.041248, CVaR -0.057676) and more stable returns compared to DDPG and MPT. performance
Source sentence
SAC, on the other hand, achieves smoother convergence with
- DDPG yields negative mean log return and higher drawdowns, indicating instability in volatile environments. negative
Source sentence
DDPG shows negative mean returns and higher drawdowns, indicating insta‑bility in volatile environments.
- The forecasting network used as an auxiliary tool attains approximately 70% prediction accuracy, reflecting market volatility and non‑stationarity. predictive-accuracy
Source sentence
prediction accuracy reaching approximately 70%.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 48 related papers). Coverage is partial, so this is neither replication nor refutation.
1 paper in the archive
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.