We present a framework for defining trading scenarios based on fish ma
rket auctions. In these scenarios, trading (buyer and seller) heteroge
neous (human and software) agents of arbitrary complexity participate
in auctions under a collection of standardized market conditions and a
re evaluated against their actual market performance. We argue that su
ch competitive situations constitute convenient problem domains in whi
ch to study issues related with agent architectures in general and age
nt-based trading strategies in particular. The proposed framework, con
ceived and implemented as an extension of FM96.5 (a Java-based version
of the Fish,market auction house), constitutes a test-bed for trading
agents in auction tournament environments, EM97.6. Finally, we illust
rate how to generate tournaments with the aid of our test-bed by defin
ing and running a very simple tournament involving a set of rudimentar
y buyer agents.