Two Stock-Trading Agents: Market Making and Technical Analysis

Abstract Evolving information te hnologies have brought omputational power and real-time fa ilities into the sto k market. Automated sto k trading draws mu h interest from both the elds of omputer s ien e and of business, sin e it promises to provide superior ability in a trading market than any individual trader. Trading strategies have been proposed and pra ti ed from the perspe tives of Arti ial Intelligen e, market making, external information feedba k, te hni al analysis et . This paper examines two automated sto k-trading agents in the ontext of the Penn-Lehman Automated Trading (PLAT) simulator [1℄, whi h is a real-time, real-data market simulator. The rst agent devises a market-making strategy exploiting market volatility without predi ting the exa t dire tion of the sto k pri e movement. The se ond agent uses te hni al analysis. It might seem natural to buy when the market is on the rise and sell when it's on the deline, but the se ond agent does exa tly the opposite. As a result, we all it the reverse strategy. The strategies used by both agents are adapted for automated trading. Both agents performed well in the PLAT live ompetition. In this paper, we analyze the performan e of these two automated trading strategies. Comparisons between them are also provided.