Hands-On Machine Learning for Algorithmic Trading

Model-free versus model-based agents

Model-based RL approaches learn a model of the environment to allow the agent to plan ahead by predicting the consequences of its actions. Such a model may be used, for example, to predict the next state and reward based on the current state and action. This is the basis for planning, that is, deciding on the best course of action by considering possible futures before they materialize.

Simpler model-free methods, in contrast, learn from trial and error. Modern RL methods span the gamut from low-level trial-and-error methods to high-level, deliberative planning, and the right approach depends on the complexity and learnability of the environment.