Parts | Days | Selection | Search | Downloads | Help
AKSOE: Arbeitskreis Physik sozio-ökonomischer Systeme
AKSOE 13: Economic Models and Evolutionary Game Theory II
AKSOE 13.2: Talk
Thursday, March 29, 2007, 12:15–12:45, H8
Efficient control in a multi-agent stock market model — •Ulrich Cherdron1, Rolf Pawelzik2, Roland Rothenstein3, and Klaus Pawelzik1 — 1Institut für Theoretische Physik, Universität Bremen, Otto-Hahn Allee 1, D-28334 Bremen. — 2Feldstr. 68, 24105 Kiel. — 356, Avenue de Cailles, 1170 Bruxelles, Belgium.
Simulations demonstrate that the exchange of stocks and money during trade in an order book market balances external influences on the return which are induced by fundamental traders. Here we present a novel formalization of order book price determination in stock markets from which we derive gradient based learning rules that optimize the dynamic parameters (stocks and money held by each agents) for control. We analyse the dynamics of both the returns and the internal dynamical parameters in a minimal model by comparing on-line versions of these learning rules with the natural adaptation caused by trading. Our results explain the emergence of control by trading which might contribute to the apparent efficiency of real stock markets.