A Python toolkit for training reinforcement learning agents and backtesting rule-based algorithms in financial markets.
TradingGym is a Python toolkit for training reinforcement learning agents and backtesting rule-based trading algorithms using historical market data. It provides a simulated trading environment where developers can test strategies, calculate rewards, and analyze transaction details before deploying them in live markets. The project is inspired by OpenAI Gym and supports both tick and OHLC data formats.
Quantitative developers, algorithmic trading researchers, and data scientists who want to prototype and test reinforcement learning or rule-based trading strategies in a controlled, simulated environment.
Developers choose TradingGym because it offers a specialized, open-source alternative to proprietary trading simulators, with a familiar Gym-like interface, support for custom data features, and plans for real-time trading integration.
Trading and Backtesting environment for training reinforcement learning agent or simple rule base algo.
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Allows configurable parameters like observation windows, step lengths, transaction fees, and feature columns, as demonstrated in the `make` function examples for tailored simulations.
Originally designed for tick data but also supports OHLC data, providing versatility for different types of market data analysis.
Inspired by OpenAI Gym, it offers a familiar reinforcement learning environment structure, making it easier for RL practitioners to adopt.
As an open-source project, it enables modification and extension for specific research needs, aligning with its philosophy to democratize algorithmic trading.
Critical sections such as training examples for DQN, policy gradient, and other algorithms are marked as WIP, lacking detailed guides or working code snippets.
Users must implement all agents and strategies from scratch, with no provided models or advanced algorithms, increasing development time.
Requires cloning the repository and running `python setup.py install`, which is less convenient than package manager installations and might lead to setup issues.