Algorithms & Strategies in Prediction Markets

Evdokia Nikolova (MIT)

Prediction markets are flourishing as a very accurate means of predicting future events (e.g., check out the Iowa Electronic Market on the current presidential election!)

Despite the use of a number of prediction market models in practice, a lot of research remains to be done on understanding the markets' strategic properties: they are continually subject to gaming and manipulation, which may distort the accuracy of their prediction. But even in the absence of manipulation, the market may fail by having to solve hard computational problems.

In this talk we describe how prediction markets work and then offer some new tools and algorithms for their analysis. Some of the results draw on surprising connections to classical combinatorial optimization problems.

This talk is based on two recent papers, "A Strategic Model for Information Markets" with Rahul Sami (EC 07) and "Betting on Permutations" with Yiling Chen, Lance Fortnow and David Pennock (EC 07)

Host: Ronitt Rubinfeld

(The Algorithms and Complexity Seminar series talks are usually held Thursdays 4:00-5:15pm in 32-G575.)