Australia ASX Gym Environment
Australia ASX Gym Environment
The Australia ASX Gym Environment is a sophisticated virtual stock exchange trading platform built on OpenAI Gym. It provides a realistic simulation environment that allows reinforcement learning agents to buy and sell stocks in real-time market conditions.
Key Features
Comprehensive Market Data
- SQLite database containing 10 years of real transaction data
- Covers approximately 2,000 ASX listed companies
- Spans 11 different market sectors
Flexible Configuration
- Simulate the entire ASX market or select specific companies of interest
- Optional brokerage fee simulation
- Configurable start dates for historical backtesting
- Weekly data update tool to maintain current market information
Multiple Render Modes The environment supports three visualization modes:
- Human Mode: Renders interactive visualizations on-screen for human observation
- RGB Array Mode: Returns RGB image arrays without on-screen rendering, ideal for video creation
- ANSI Mode: Displays text-based output on the screen without images
Data and Visualization
- Utilizes mpl_finance library for professional stock data rendering
- Automatically saves episode data (daily portfolio values) for further analysis
- Generates and stores images at each simulation step
- Creates video-ready image sequences in human and RGB array modes
Simulation Output
During each training episode, the ASX Gym Environment generates:
- Step-by-step portfolio value tracking
- Visual representations of market conditions and trading activity
- Episode summary data for performance analysis
- Image sequences suitable for creating training videos