Overview
TradingGym is a Python-based toolkit designed to facilitate the training and backtesting of reinforcement learning (RL) agents and rule-based algorithms for trading. Inspired by OpenAI Gym, it provides a framework for creating custom trading environments with tick data or OHLC data formats. The architecture supports the simulation of trading scenarios with features like transaction fees and position limits. It allows users to define feature columns, such as price and volume, to be used as inputs for trading status. The environment integrates with brokerage APIs for real-time trading in the future. It offers functionalities to analyze transaction details, making it suitable for developing and testing various trading strategies including DQN, policy gradient, actor-critic, and A3C with RNN.