My original contribution to knowledge is to evaluate the performance of adaptive trading agents for the continuous double auction (CDA) under realistic experimental conditions. Autonomous trading agents are a significant application of agent-based computational economics (ACE), the common ground between multiagent systems and economics; the CDA is arguably one of the most popular economic institutions, for the high efficiency that it offers, and for its widespread use in real world financial securities exchanges. ACE researchers proposed several trading agents for the CDA, the most prominent of which are periodically upgraded: there is an ongoing informal competition among them, to determine which strategy performs best. The creators of those agents often refer to the potential relevance of their work to real world financial markets; and yet although much interest has been given to the behavioural details of the various trading strategies, the conditions adopted for experimentation are too often far from those in place in the real world of the financial markets industry. My central question is: to what extent are those software trading agents applicable to real world financial markets? The aim of this study is to measure the performance of those software trading agents under experimental conditions that resemble those of real world markets; for that purpose, I use both pure computational agent markets, and mixed markets of human and software trading agents. To recreate the experimental environment of real world markets: I consider the rules of major stock exchanges; I draw inspiration from state-of-the-art experimental economics models; and I create a two-sided trading agent capable of improving the "quality" of the market. My main findings are that under realistic experimental conditions, the selected software trading agents form a highly efficient market, whose performance is only marginally reduced by the more stringent constraints I added; and in mixed markets, software trading agents outperform humans.
Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:683918 |
Date | January 2015 |
Creators | De Luca, Marco |
Publisher | University of Bristol |
Source Sets | Ethos UK |
Detected Language | English |
Type | Electronic Thesis or Dissertation |
Page generated in 0.0138 seconds