Card games pose an attractive opportunity for per- forming experimentation, along with uncertainties and the partial observability of the game with reinforcement learning.This ap- proach aims to discover the most effective strategies for winning consistently in the specific card game of Blackjack.So its obvious that people playing these games involving money would like to explore new strategies to increase winnings or minimize losses. This paper explores model-free methods such as Q Learning and SARSA[1], seeking the optimal policy for playing Blackjack against a dealer.It tries to find the policies, which indicate the best action based on the player’s score and the dealer’s visible card. The observed policy aligns with the fact how players with advanced techniques apply specific strategies in their games.

read more about this project: https://github.com/rednayan/Adaptive-Blackjack-Strategies/blob/master/paper.pdf