Artificial Intelligence of Modern Board Games

Student: Khayyam Zubair
Mentor: Sean McCulloch (Department of Mathematics and Computer Science)

The goal of our research was to build a computer program to play a game of Modern Art. Modern Art is a modern board game that is based on auctions. We aimed to use probability based mathematics to form in game strategies for our program. A truly artificially intelligent program would play the game as well as a human being could.

Modern Art is an auction based board game about the world of art galleries. The goal of our work was to come up with an Artificially Intelligent computer   program to play a simulated version of Modern Art. Modern Art is played by 3-5 players acting as gallery owners. Each player is dealt some cash and a hand of cards, which are paintings that  the players will auction to one another. Any income from selling these paintings goes to the selling player. Each painting bought during such an auction is sold at the end of a round to the bank and the buyer’s compensation is based on how well the artist of the painting had performed in that round. Our work on the project focused in two directions: developing a computer simulation of the game and developing a winning strategy that a computer player could use to play the game and challenge a human competitor.

Simulating this game with code required coming up with models and simulations for cards, artists and auctions. The game logic kept track of how much money each player made and how well each artist performed. Winning in modern art hinges on bidding the right amount on a painting. A player has to decide what a painting’s future worth but also the net value of their competitor after the transaction. We hypothesized that such a prediction could be made based on probability. The AI game engine was thus tasked to randomly simulate a game at the given state a number of times and then rank the performance of that artist in each of these simulations, then calculate a probable worth based on these results. We found that the program was able to read the course of a game and make reasonable bids for the paintings up for auction.