File:Flea Market Simulator- A Market Simulator for Experiments on the Computational Evolution of Trading Strategies for Economic Agents.pdf

In this paper, we present the Flea Market Simulator: a minimal market simulator, where agents have strategies they use to determine which goods to trade and at what rate. The primary goal is to understand how intelligent strategies can develop in a dynamic economic environment. The simulator is designed to run multiple independent strategy generating functions at once, for isolated sub-populations of agents in the market. The timing and extent of interaction between these sub-populations can be flexibly controlled by the user. Resultant differences between subpopulations provides data with which to compare the different strategy generating functions. Specifically, these strategy generating functions are intended to be evolutionary computational (EC) methods, and as such, the simulator provides the framework for tournaments between EC methods, in terms of their ability to produce “successful” agent trading strategies. The definition of success is left purposefully open-ended.