Trading agents have to solve combinatorial assignment problems to perform well

By a trading agent we mean software that trades combinations of goods on multiple markets on behalf of one or more clients, based on expressed or elicited client preferences. A software trading agent can potentially handle more complex combinations of goods, larger numbers of goods and markets, a wider range of market types, and faster markets with more fine-grained goods than its human counterparts.

Relative to their clients, trading agents have to solve a combinatorial assignment problem, where goods available to an agent are packaged into bundles, and delivered as such to the clients. This problem is related to that of an auctioneer determining winning bids in a combinatorial auction. A successful trading agent must therefore make use of similar integer programming techniques or, alternatively, techniques of constraint programming.

Aspects of the trading agent problem