Computer Science Students Develop Blackjack Buddy
January 8th, 2013 by Site Admin
The students who had enrolled for Associate Professor Fred Martin’s Artificial Intelligence (AI) Course developed some interesting projects, one of which was Blackjack Buddy, an AI agent that can advise blackjack players about the best move to make while playing blackjack.
Blackjack Buddy was developed by Jonathan Yu, one of the computer science students who had taken up the course, with the intention of designing and implementing an artificial intelligence agent, which makes use of Q-Learning to help players effectively play blackjack even if they do not know the rules and strategies of the game.
Speaking about his project, Yu said that the agent acts as an advisor and suggests the best moves a human blackjack player can make at the gaming table. Besides, the agent learns continuously during the game. He said that the project, which required not only creating AI methods, but also learning the game of blackjack, was challenging as well as rewarding.
Yu also said that Blackjack Buddy is capable of learning one optimal strategy per 5,000 simulations of blackjack, approximately. While the win rate of random strategies is around 25 – 30 percent, the win rate of Blackjack Buddy is around 40 – 50 percent, in line with the best and most talented human players.
Leave a Reply