Reinforcement Learning: Learning How to count cards at Blackjack

Introduction

In this project I use RL to make an agent learn the optimal blackjack strategy to play. The optimal action to play is given by the Bellman equation

In a second time it is demonstrated that the agent can learn to leverage a simple card counting strategy to improve its games and beat the casino.

Instructions

Run the script Environment.R to launch the game.