|
CS Home
Dept Info/Contacts
People
Research
Events
Courses
Undergrad Programs:
Computer Science
Informatics
Graduate Program
Prospective Students
Faculty Hiring
Employment
Resources
Help Lab Hours
Student Groups
Support the Department: Weeg Professorship
|
|
Strategic Behavior and Combinatorial Betting in Prediction
Wednesday, March 5, 2008
3:30-5:000pm, W401 PBB
Abstract Situated in a de facto standard market maker
mechanism, logarithmic market scoring rules, we study prediction
markets from both economic and computational perspectives. From the
economic perspective, we investigate the equilibrium behavior of
informed traders in prediction markets. We examine what information
structures lead to truthful play by traders, meaning that traders
reveal all of their information honestly as soon as they are able. We
show that when signals of traders are independent conditional on the
state of the world, truthful betting is a Perfect Bayesian Equilibrium
(PBE). However, when signals are conditionally dependent, there exist
joint probability distributions on signals such that at a PBE traders
have an incentive to bet against their own information - strategically
misleading other traders in order to later profit by correcting their
errors. From the computational perspective, we design expressive
betting languages for combinatorial prediction markets and examine the
computational problem of pricing such markets. In our combinatorial
markets, traders are allowed to submit bets of the form "horse A
finishes in position 1, 2, or 5", "horse C beats horse D", "Hillary
Clinton wins Ohio and Florida", or "Duke wins a third round game in
the NCAA basketball tournament". Pricing such markets are
computationally intractable except for the single-elimination
tournament betting, in which case, we use a Bayesian network to
facilitate price updates. This is the first example of a tractable
market-maker driven combinatorial market.
|
