Computational Analysis of Position Auctions

Members

Kevin Leyton-Brown
Samantha Leung
David Robert Martin Thompson (Contact at: )

Funding

This work was supported by a grant under Microsoft's "Beyond Search" program:

Title: Equilibrium Computation and Semi-automated Mechanism Design for adCenter Auctions
Principal Investigator: Kevin Leyton-Brown

Abstract: The proposed research will use a recently developed compact representation for game theory ('action-graph games') to compute equilibria of realistic ad auctions. The researchers will mine the released adCenter data logs to develop a realistic model. Given a game-theoretic model, it is possible to perform "semi-automated mechanism design," a computational process that will recommend alternate auction designs that would yield improved performance according to a given benchmark. The researchers will freely release software for building and analyzing such models, and publish the research results.

Media:

2008 Poster
Slides from our presentation at the 2009 Beyond Search Workshop

Data

Outcomes.csv v1.0 - Data from the EC'09 paper

Code

Position Auctions to AGGs v0.9 - Code used for the EC'09 paper, needs cleanup / documentation
[available under the "MIT License"]
Software for manipulating, visualizing and solving AGGs can be found at agg.cs.ubc.ca.