The Trading Agent Competition

Brown University, Providence RI

Investigators

Abstract

This is a conference-related grant to support the travel, subsistence and registration expenses of approximately 16 student participants in a trading agent competition, being held in Hakodate, Japan, plus providing computational resources to enable educational experiences. The Trading Agent Competition (TAC) is an international forum designed to promote and encourage high-quality research about trading agents. TAC provides a platform for researchers to evaluate programmed trading techniques by competing with agents from other design groups in a simulated market scenario. TAC tournaments have been held annually since 2000, and have attracted participants from institutions in dozens of countries around the world. Involvement of American students will not only advance their individual careers in computer science, but will also contribute to the nation's future science and technology workforce. This activity contributes to curriculum development by developing and demonstrating methods for involving students in the design and competitive use of artificial agents. Entries in the Trading Agent Competition are software programs designed to trade in electronic markets. They are called "agents" because these programs operate autonomously in the market: sending bids, requesting quotes, accepting offers, and generally negotiating deals according to market rules. Although the agent's activity is ultimately determined by its programmers, the trading behavior is itself fully automated: humans do not intervene while the negotiation is in progress. Trading agents face a most challenging task. To play the market effectively, an agent must make real-time decisions in an uncertain and fast-changing environment, taking into account the actions of other agents that are doing the same. Capable agents rapidly assimilate market information from many sources, forecast future events, optimize complex offers and resource allocations, anticipate strategic interactions, and learn from experience. Successful trading agents adopt and extend state-of-the-art techniques from artificial intelligence, operations research, statistics, and other relevant fields. The annual trading agent competitions were initiated to promote research and education in the technology underlying trading agents. At the annual competition, the developers of techniques in trading strategy evaluate these ideas and communicate their results in a public forum for the benefit of the broader research community. The educational function of TAC is manifest by the significant role of students on almost all teams entering the competition. Many universities use TAC as an exercise for teaching about electronic commerce and artificial intelligence techniques. TAC is a useful aid in the classroom because it is by nature a hands-on (laboratory-like) experience for computer scientists. Research and education in trading agents promises to improve the state of art and practice in their development, and ultimately lead to more effective electronic markets. Equally important, increasing public knowledge in this area promotes understanding of the behavior of autonomous software agents as such systems become more prevalent in commerce and other domains.