In Ad Auctions, agents participate in a simplified sponsored search scenario. The ad publisher, as well as a user population that searches and purchases, are all simulated by the game server. The participating agents play the role of advertisers that bid for placement of their ads in ranked lists viewed by the querying users in the result pages. User preferences regarding the products and the state of each user are not always fully revealed, while both affect the probabilities of clicking and converting. When a user clicks on an ad, the advertising agent is charged a cost per click. If that click leads to a conversion, i.e. the user finally purchases the advertised product, the advertiser earns profit. The storage capacity of each agent adds an aditional constraining factor.

The advertising agents receive daily feedback reports from the publisher that reveals only a portion of information for other advertiser activity. Finally, given some prior probabilities about user behaviour and through analysis of the limited information, agents must bid the optimal amount for ad placement in order to maximize conversions and minimize clicks made.

More detailed information can be found here.