tailieunhanh - Development and testing of an artificial stock market

Third, this paper contributes detailed specifications of promising strategy designs, one of which vastly outperformed competitor strategies in an open stock-trading competition and exhibited consistent profitability under a variety of market conditions. The remainder of the paper is organized as follows. Section 2 provides the relevant technical background on the PXS simulator, our substrate domain. Section 3 charac- terizes prior research and points out the distinguishing features of this work. Section 4 discusses our approach to the automated stock-trading problem, explains our assump- tions, and details our experimental methodology. Sections 5–7 describe our three stock- trading strategies. Sections 8–10 present and analyze the experimental results, focusing, respectively, on individual evaluations,. | Development and testing of an artificial stock market Michele Marchesi Dipartimento di Ingegneria Elettrica ed Elettronica Universita di Cagliari Piazza d Armi 09123 Cagliari Italy. Silvano Cincotti Dipartimento di Ingegneria Biofisica ed Elettronica Universita di Genova Via Opera Pia 11a 16145 Genova Italy. Sergio Focardi Intertek Group rue De Javel 94 75015 Paris France Marco Raberto Dipartimento di Ingegneria Biofisica ed Elettronica Universita di Genova Via Opera Pia 11a 16145 Genova Italy. September 15 2000 Abstract In this paper an artificial financial market based on heterogeneous agents is presented. The proposed market is composed of traders with limited amount of cash one traded asset and a centralized mechanism the market maker matching buy and sell orders. The price formation process is given by the intersection of the demand and the supply curve. The artificial financial market has been implemented using advanced software engineering techniques in particular extreme programming and object oriented technology. The resulting system is a To whom correspondence should be addressed. Email michele@ 1 2 Modelli Dinamici in Economia e Finanza - Urbino 2000 powerful tool able to numerically simulate financial market operations in a realistic way. Preliminary results show that the price time series exhibits a random walk behavior with fat tails distribution of returns. 1 Introduction In the last decade many computer simulated artificial financial market have been proposed 1 2 3 4 5 . For a review see LeBaron 6 . Perhaps the most famous among them is the artificial market - populated with heterogeneous agents endowed with learning and optimisation capabilities - built at Santa Fe Institute by Arthur Holland et al. 7 8 9 . In this paper we present an agent-based computer simulator of a stock market called the Genoa artificial market in honour of to the beautiful city where most of this work was performed. In the Middle Age Genoa was a major financial