Analyzing Unique-Bid Auction Sites for Fun and Profit

Unique-Bid auction sites are gaining popularity on the Internet in recent years. We have managed to extract dynamic temporal bidding data from such a site, using a backpropagation algorithm for analysis of side signals. This offered us rare insights on actual bidding strategies used by actual bidders, such as bidding-bursts, late-bidding and position-targeted bidding. We constructed an agent-based model simulating these behaviors, and validated it using the extracted bidding data. This model allowed us to experiment with different strategies of our own. We devised a set of automated winning strategies that performed well on our simulated environment. Finally, we demonstrated some of our strategies against a commercial auction site, achieving a 91% win rate and over 1000 UK pounds profit.