Overview
gym-anytrading is an open-source project providing a collection of OpenAI Gym environments specifically designed for developing and testing reinforcement learning (RL) based trading algorithms. It aims to simplify and improve the process of creating and evaluating these algorithms, primarily for FOREX and stock markets. The core of the library revolves around three key environments: TradingEnv, ForexEnv, and StocksEnv. TradingEnv serves as an abstract base class, offering a general-purpose framework. ForexEnv and StocksEnv extend TradingEnv, adding market-specific features. The environment simplifies action spaces (Buy/Sell) and positions (Long/Short) to improve RL agent learning efficiency. It uses pandas DataFrames for data input, allowing flexible feature engineering via the _process_data abstract method.