We present a framework for defining trading scenarios based on fish market auctions: Trading (buyer and seller) heterogeneous (human and software) agents of arbitrary complexity participate in auctions under a collection of standardized market conditions and are evaluated against their actual market performance. We argue that such competitive situations constitute convenient problem domains in which to study issues related with agent architectures in general and agent-based trading strategies in particular. The proposed framework, conceived and implemented as an extension of FM96.5 (a Java-based version of the Fishmarket auction house), constitutes a testbed for trading agents in auction tournament environments. We illustrate
how to generate tournaments with the aid of our testbed by defining and running a very simple tournament involving a set of buyer agents developed by undergraduate students from Ecole Polythechnique Fédérale de Lausanne (EPFL).