Monte Carlo Methods — Learning from Complete Episodes

Dynamic programming needs a model. Monte Carlo needs nothing but episodes. Run the policy, watch the returns, average them. The simplest idea in RL — and the one that unlocks everything downstream.