Performance Analysis of a Counterintuitive Automated StockTrading Agent.pdfVIP

Performance Analysis of a Counter-intuitive Automated Stock-Trading Agent Ronggang Yu and Peter Stone Department of Computer Sciences The University of Texas at Austin /~{ryu,pstone} ABSTRACT uses real-world, real-time stock market data available over Autonomous trading in stock markets is an area of great modern ECNs. It is the first simulator to incorporate com- interest in both academic and commercial circles. A lot of plete order book information, thereby allowing it to “match” trading strategies have been proposed and practiced from agent orders with real-world orders and simulate the result- the perspectives of Artificial Intelligence, market making, ing effects on the market. It also provides a lot of APIs so external data indication, technical analysis, etc. This paper that the participants can program their strategies and trade examines some properties of a counter-intuitive automated with other agents and outside markets automatically. stock-trading strategy in the context of the Penn-Lehman Automated Trading (PLAT) simulator [1], which is a real- Using these and other simulators, as well as experiments time, real-data market simulator. While it might seem nat- in the real world, many researchers have studied trading ural to buy when the market is on the rise and sell when it strategies from the perspectives of Artificial Intelligence [7, is on the decline, our strategy does exactly the opposite. As 8, 11], neural networks [9], technical analysis [10], etc., How- a result, we call it the