Implementing a 9.2 Quintillion Outcome Prediction Market
Speaker: David Pennock, Yahoo! Research
Location: Warren Weaver Hall 1302
Date: November 19, 2010, 11:30 a.m.
Host: Mehryar Mohri
I'll describe our efforts at Yahoo! Research to build a true combinatorial prediction market with exponentially many outcomes -- too many to store and price explicitly. Our market for the men's college basketball playoffs had 2 to the power 63 or 9.2 quintillion outcomes representing every possible way the tournament could unfold. Our market maker can price 2 to the power 2 to the power 63 bets (more than a gogol but less than a googolplex), yet has constant bounded loss. The market maker runs an implicit version of Hanson's logarithmic market scoring rule using importance sampling. A beta version called Predictalot (predictalot.yahoo.com) is running on the Yahoo! Application Platform for anyone to install and use. In the game, you can predict almost anything about a sports tournament, like "Germany will advance further than both Brazil and France" in the World Cup, or "the sum of seeds of the final four teams will be a prime number". You can compose any of millions of predictions and sell them any time for virtual points, even in the middle of a match, just like the stock market. Predictalot is also a massive crowdsourcing experiment. Our goal is to take the wisdom of crowds to the extreme by computing what all these different predictions, taken together, say about the likelihood of each of the trillions or more possible ways the tournament could unfold.
David Pennock is a Principal Research Scientist at Yahoo! Research in New York City, where he leads a group focused on algorithmic economics. Pennockâ€™s job is to dream up technologies that are disruptive and revolutionary, either commercially, scientifically, or both. He has over fifty publications relating to electronic commerce and the web, including papers in PNAS, Science, IEEE Computer, Theoretical Computer Science, AAAI, EC, and WWW. He has given over forty talks and authored two patents and ten patent applications. In 2005, he was named to MIT Technology Reviewâ€˜s list of 35 top technology innovators under age 35. Pennock is part of a growing vanguard of computer scientists and economists who are working together to investigate the role of computation in economic theory and to design and build the marketplaces of the digital age. One of his primary areas of expertise is the design and analysis of prediction markets.
Refreshments will be offered starting 15 minutes prior to the scheduled start of the talk.