My current research interest is called "Artificial Intelligence for Modern Board Games". I typically work with students on this project in summer research. Here are the games we have worked on in past summers:
- Football Strategy is a representation of a football game where the offense and defense pick plays, and the chart that determines the outcome can be viewed as a Normal Form game, and thus can be analyzed by Game Theoretic methods. I developed a solution to the Game Theoretic aspects to the game, which creates a mixed strategy of plays to call, both on offense and defense. I also have come up with ways to adjust the chart to take into account more strategic information (the score, the time remaining in the game, field position, and so on). [Sean McCulloch, "A Game Theoretic Intelligent Agent for the Board Game Football Strategy"; MAICS 2015.]
- Battle Line is a card game where players develop a three-card poker hand simultaneously over nine different “flags” (Each flag is one set of three cards for each player. On a player’s turn, they can play a card on any of the nine flags). The underlying ideas for the AI for this game are probability-based (for example, if you’re waiting to draw a card to make a hand, we can determine the probability of you drawing that card) and we can use this probability to design general strategies for the game. [Sean McCulloch, Daniel Bladow*, Tom Dobrow*, Haleigh Wright*. “Deep Barca: A Probabilistic Agent to Play the Game Battle Line”; MAICS 2017]
- Modern Art is a multi-player game where players bid in a variety of different ways on pieces of art that other players put up for auction. By deciding what art to sell, and how much to bid, the players also determine the final selling price for the piece of art. The “optimal bid” that should be made can be computed mathematically, but it is dependent on the estimate of the final selling price, which needs to be estimated by each player.
- Pandemic is a cooperative game where players work together to treat disease outbreaks around the world, and collect cards that will cure the diseases. Our work on this game involved writing the basic skeleton of the game, and a set of “tools” that could help a player estimate what the best moves would be (for example, the expected number of turns in the game until the game is lost or won).
- Eurorails is a network-based game where players design a rail network (in the board game, this is done by physically drawing on the board in crayon) to pick up goods from cities in Europe, and deliver them to other cities in Europe. There is a lot of interesting Graph Theory in this game, in determining the best way to create and traverse a network of cities. Our students wrote a program that managed the rules of the game (including a very idiosyncratic map) and then added a very nice graphical client, along with some advanced search algorithms, and the agent now plays a very strong game.
- Scotland Yard is a semi-cooperative game where a team of detectives work together to close in and capture “Mr. X”, who moves secretly, and only gives partial information about his position to the detectives. There is a very interesting tree-building mechanic, where the detectives build a tree of possible locations of Mr. X, and then decide which places to investigate to reduce the degree of the tree.