RL Crypto Trading Bot

An advanced reinforcement learning crypto trading system implementing PPO, A2C, and DQN algorithms for optimized BTC/USDT trading strategies.

📌 Project Highlights

Why RL for Crypto Trading?

• Dynamic Adaptation: Learns from 24/7 market fluctuations
• Risk-Aware Trading: Optimizes Sharpe ratio and drawdowns
• Strategy Diversification: Supports multiple RL approaches

Algorithm Performance Summary

Algorithm	Avg Reward	Sharpe Ratio	Returns (%)	Key Strength
DQN	184.69	2.22	1.28	Best Risk-Adjusted Returns
PPO	173.65	1.85	0.97	Balanced Performance
A2C	165.78	1.82	0.99	Fast Convergence

🚀 Core Features

• Multi-algorithm framework (PPO/A2C/DQN)
• Real-time trading visualization
• Advanced metrics tracking:
  • Portfolio volatility (0.01-0.05)
  • Win rate analysis (51.76%-51.90%)
  • Drawdown monitoring (-0.24% to -2.31%)
• Buy & Hold benchmarking

📊 Performance Analysis (BTC/USDT)

Comparative Metrics

Metric	DQN	PPO	A2C	Buy & Hold
Avg Reward	184.69	173.65	165.78	-
Returns (%)	1.28	0.97	0.99	0.13
Sharpe Ratio	2.22	1.85	1.82	2.20
Max Drawdown	-2.31	-1.94	-1.99	-0.24

Key Observations

DQN Dominance:

• Highest returns (1.28%) and Sharpe ratio (2.22)
• Competitive win rate (51.90%)

Buy & Hold:

• Lowest volatility (0.01)
• Surprisingly high Sharpe ratio (2.20)

All RL Models:

• Consistently outperform Buy & Hold in returns
• Maintain >51.7% win rates

⚙️ Setup & Usage

Installation

git clone https://github.com/Vnadh/RL-Crypto-Trading-Bot.git

cd RL-Crypto-Trading-Bot

python -m venv myenv

source myenv/bin/activate  

pip install -r requirements.txt

Training Models

# Train all models with optimized parameters
python train_model.py 

Performance Evaluation

# Generate detailed metrics report
python testing_model.py

📈 Strategic Insights

1. DQN Superiority: 42% higher returns than Buy & Hold
2. Risk Management: All RL models maintain Sharpe ratios >1.8
3. Market Resilience: PPO shows most stable drawdown profile

🌟 Unique Advantages

• Flexible Action Spaces: Supports both discrete and continuous trading
• Live Rendering: Watch agents trade in real-time
• Parameter Optimization: Pre-tuned configurations for crypto markets

🤝 Contribution Guidelines

1. Fork the repository
2. Create feature branches
3. Submit PRs with test results
4. Maintain coding standards

📜 License

MIT Licensed - See LICENSE for details.

---

Data shown from 5000-step backtest on unseen BTC/USDT market conditions