Wyckoff Trading Assistant

A comprehensive web application that combines Reinforcement Learning (RL) with Wyckoff methodology principles to assist traders in analyzing stocks and making informed trading decisions.

Overview

Wyckoff Trading Assistant offers two main features:

1. Wyckoff AI Assistant: Ask questions about Wyckoff methodology, market structure, and trading principles
2. RL-Enhanced Trading Strategy: Backtest a Q-learning based trading strategy on any stock

The application uses a reinforcement learning algorithm to learn optimal trading strategies based on historical price data, combined with an AI chatbot trained on Wyckoff methodology principles.

Features

Wyckoff AI Assistant

• Interactive chat interface for questions about Wyckoff methodology
• Trained on comprehensive Wyckoff trading principles and concepts
• Answers questions about springs, upthrusts, accumulation, distribution, and more
• Helps traders understand market structure and price action

RL Trading Strategy

• Q-learning based trading algorithm that learns optimal Buy/Sell/Hold decisions
• Customizable backtesting on any stock symbol
• Performance metrics including ROI, portfolio value, and trade signals
• Visual representation of trading signals overlaid on price charts

Interactive Dashboard

• Real-time stock data visualization
• Technical indicators (MA20, MA50, MA200, RSI, Bollinger Bands)
• Performance metrics for strategy evaluation
• Trade signal visualization

Technology Stack

• Backend: Flask (Python)
• Frontend: HTML, CSS, JavaScript, Chart.js
• Machine Learning: PyTorch (for the Wyckoff chatbot), Q-learning (for the trading strategy)
• Data: yfinance for real-time and historical stock data

Installation

1. Clone the repository:

git clone https://github.com/yourusername/wyckoff-trading-assistant.git
cd wyckoff-trading-assistant

2. Create and activate a virtual environment:

python -m venv wyckoff_venv
# On Windows
wyckoff_venv\Scripts\activate
# On macOS/Linux
source wyckoff_venv/bin/activate

3. Install dependencies:

pip install -r requirements.txt

You can request the model file from here

4. Make sure you have the model file:

• Place your transformer model file (transformer_chatbot_gpu_deco_2.pth) in the correct location
• Update the path in app.py if needed

5. Run the application:

python app.py

6. Open your browser and navigate to:

http://localhost:5000

Reinforcement Learning Trading Strategy

The trading strategy uses Q-learning, a model-free reinforcement learning algorithm that learns a policy to maximize expected future rewards. The implementation includes:

Environment (StockTradingEnv)

• State: Current stock price and position (holding/not holding)
• Actions: Buy, Sell, Hold
• Rewards: Changes in portfolio value

Q-Learning Algorithm

• Q-Table: Stores state-action values
• Epsilon-Greedy Policy: Balances exploration and exploitation
• Learning Rate (Alpha): Controls adaptation speed
• Discount Factor (Gamma): Weighs importance of future rewards

State Discretization

• Continuous price values are mapped to discrete bins
• Enables tabular representation for the Q-learning algorithm

Backtest Procedure

1. Historical data is fetched for the specified stock and time period
2. The environment is initialized with the data
3. Q-learning algorithm trains on the data to learn optimal policy
4. The learned policy is evaluated on the same data for performance metrics

Wyckoff Methodology Integration

The application integrates key Wyckoff concepts:

1. Price Action Analysis: The RL model learns to recognize favorable price patterns
2. Volume Analysis: Trading decisions consider volume as a confirming factor
3. Wyckoff Phases: The AI assistant helps identify accumulation, markup, distribution, and markdown phases
4. Key Events: Springs, upthrusts, and tests are explained by the AI assistant and potentially identified in the price data

Project Structure

wyckoff-trading-assistant/
├── app.py                 # Main Flask application
├── trading_strategy.py    # RL trading algorithm
├── requirements.txt       # Python dependencies
├── static/
│   ├── css/
│   │   └── style.css      # Application styling
│   └── js/
│       └── main.js        # Frontend JavaScript logic
├── templates/
│   ├── index.html         # Main application interface
│   ├── docs.html          # Documentation page
│   ├── 404.html           # Error page
│   └── 500.html           # Server error page
└── utils/
    ├── __init__.py
    ├── data_processing.py # Data handling functions
    └── model_handler.py   # Wyckoff model utilities

Usage

Dashboard

1. Enter a stock symbol in the search bar and click "Load"
2. View the stock price chart with technical indicators
3. Click "Run Backtest" to test the RL trading strategy

Wyckoff Assistant

1. Navigate to the Wyckoff Assistant tab
2. Type your question about Wyckoff methodology or trading principles
3. View the AI-generated response based on Wyckoff principles

Backtest Settings

1. Configure custom parameters (stock symbol, date range, initial capital)
2. Run the backtest with your settings
3. View detailed performance metrics and trading signals

Future Enhancements

• Portfolio backtesting with multiple assets
• Custom technical indicators and trading rules
• Advanced Wyckoff pattern recognition
• Real-time trading signals and alerts
• Enhanced visualization of Wyckoff market phases
• User accounts to save preferences and backtest history

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgements

• The Wyckoff methodology originated from Richard D. Wyckoff's work in the early 20th century
• Q-learning algorithm developed by Watkins and Dayan (1992)
• Special thanks to the open-source communities behind Flask, PyTorch, and Chart.js

---