Architecture of a prediction economy

Abstract

A design and implementation of a Prediction Economy is presented and compared to alternative designs. A Prediction Economy is composed of prediction markets, market managers, information brokers and automated trading agents. Two important goals of a Prediction Economy are to improve liquidity and information dispersal. Market managers automatically open and close appropriate markets, quickly giving traders access to the latest claims. Information brokers deliver parsed data to the trading agents. The agents execute trades on markets that might not otherwise have much trading action. Some preliminary results from a running Prediction Economy are presented, with binary markets based on football plays during a college football game. The most accurate agent chose to enter 8 of 32 markets, and was able to predict 7 of the 8 football play attempts correctly. Source code for the newly implemented tools is available, as are references to the existing open source tools used.