Price Predicting Trading Strategies for Simultaneous One-Shot Auctions
Add to Google Calendar
1. Abstract if you care about auctions/trading
We study a canonical form of simultaneous markets, where auctions operate according to sealed-bid rules. Bidding strategies for simultaneous interdependent markets typically rely on price predictions to manage uncertainty across markets. We show that optimal trading strategies in this setting can be characterized without loss of generality as price-predicting bidders, and that Bayes-Nash equilibria correspond exactly to self-confirming price predictors. Employing approximately correct predictions entails only bounded loss, and approximately self-confirming price predictors are approximately in strategic equilibrium. We demonstrate the usefulness of these concepts by developing procedures to search for price predictions that are self-confirming with respect to implementable bidding strategies. A comprehensive empirical game-theoretic analysis encompassing a broad portfolio of heuristic strategies finds strong support for this approach across bidding environments.
2. Abstract if you do not care about auctions/trading per se
Trading in simultaneous interdependent markets represents a challenging problem for autonomous agents, posing problems in optimization under uncertainty and accounting for strategic interactions with other agents. Despite the conceptual simplicity of the simultaneous auction game, to date it has eluded successful game-theoretic analysis by auction theorists. Our investigation of this problem is a case study in exploiting insights from agent architecture and computational experimentation to a rich and practically relevant multiagent scenario. The exercise illustrates in particular the possibility of proving an architectural element correct (in a useful sense), and techniques for experimentally teasing apart the effects of multiple agent strategy elements despite interactions with each other and the strategies of other agents.
Michael P. Wellman is Professor of Computer Science & Engineering at the University of Michigan. He received a PhD from the Massachusetts Institute of Technology in 1988 for his work in qualitative probabilistic reasoning and decision-theoretic planning. From 1988 to 1992, Wellman conducted research in these areas at the USAF's Wright Laboratory. For the past 19+ years, his research has focused on computational market mechanisms for distributed decision making and electronic commerce. As Chief Market Technologist for TradingDynamics, Inc. (now part of Ariba), he designed configurable auction technology for dynamic business-to-business commerce. Wellman previously served as Chair of the ACM Special Interest Group on Electronic Commerce (SIGecom), and as Executive Editor of the Journal of Artificial Intelligence Research. He is a Fellow of the Association for the Advancement of Artificial Intelligence and the Association for Computing Machinery.