Easy21 Reinforcement Learning
Tabular RL agents for the Easy21 card game (David Silver's RL course)
This project implements the Easy21 card game environment as specified in David Silver’s Reinforcement Learning Course final project.
Agents implemented:
- Monte Carlo Control — tabular, model-free policy optimization using episode returns
- TD Learning (Sarsa) — online temporal-difference updates
- Sarsa(λ) — eligibility traces with linear function approximation for generalization across states
The project explores how different RL algorithms balance exploration vs. exploitation and how function approximation can scale learning beyond tabular settings.
GitHub: ruyi101/Easy21-Game-RL