In writing our initial proposal we all agreed that we wanted to explore combinatorial prediction markets. It fell to me to explore possible application areas, and after looking into a half dozen possibilities, I recommended international military instability. This area met several criteria, including the key criteria of asking questions where better answers are worth a whole lot to the Department of Defense. During Phase I, Dave Porter and I designed and ran the Phase I lab experiments on manipulation, which found that incentives to manipulate had little or no measurable effect on price accuracy. I also ran several meetings with Al Goldman and DC area intelligence experts (the only such PAM meetings), to see what sorts of markets they would prefer.
Just before Phase I commenced, I had designed and implemented a combinatorial market maker, and proposed that PAM use it. Charles Polk, president of Net Exchange, chose instead to refit their existing software to this application. After the simulation failure at the end of Phase I, I threatened to quit. John Ledyard, CEO and founder of the firm, then got more involved in the project and proposed we run a laboratory horse race, between standard markets, my market maker, and a new combinatorial call market John would design. I designed the experimental environment in which the mechanisms were compared, Net Exchange created new software to run the mechanisms, and John ran the experiments with Caltech students.
For my work, I received a total of about $30,000 of the DARPA PAM funding. A group of us at GMU won a small FutureMAP grant in 2003, but had officially spent none of it when FutureMAP was cancelled.