Prediction Markets = Junk Science???
Chris F. Masse October 12th, 2006
The sarcasm comes from the highly opinionated British site, “The Register” (which our good doctor Robin Hanson seems to read):
Even the best-read prophet of “collective intelligence”, journalist James Surowiecki, cautions against a hype he helped create. A fan of prediction markets, Surowiecki discovered only a rare alignment of circumstances could ever provide something useful, and he hedged around the tricky subject of gaming and capture - perennial problems for Google and Wikipedia respectively.
The news of the day: MIT launches Center for Collective Intelligence -
Here’s how the MIT CCI site introduces prediction markets:
How can groups of people make accurate predictions of future events? For instance, in prediction markets, people buy and sell predictions about uncertain future events, and the prices that emerge in these markets are often better predictors than opinion polls or individual experts. When and how do these prediction markets work best? How can they be combined with simulations, neural nets, and other techniques?
If you have nothing to do tomorrow, Friday, October 13, 2006, the MIT CCI will web cast live about the launch of a collective book project (a la Wikipedia), where hundreds or even thousands of ignorants will have their say on topics they never studied in depth. (If Justin Wolfers of the Wharton School of Business, one of the contributing entities, happens to be part of the writing team, then I’ll publish a public apology.)
My Remark: I don’t see any prediction market luminaries on the MIT CCI people webpage. On the other hand, I see 90% of the scholars and industry players listed on the Midas Oracle authors webpage. (For your information, I’ve recently added some scientists and professors, and some big exchange is toying its participation.) The collective blog mechanism is better than the wiki model for idea discovery and for fostering good debates. Wikis seems to be great at publishing common-denominator explainers (if you can see what I’m trying to convey).
How Would You Answer The Two Questions? Prediction markets work best with real money, when traders trust the judgment mechanism, and when they can use Web-based resources that publish advanced indicators. As for combining prediction markets with other techniques, it’s stupid. Prediction markets are the ultimate aggregation mechanism. Period.
Addendum: Chris Hibbert’s previous blog post on the MIT CCI. (And see his anti-sarcastic comment just below this post.)
Could one create a predictability metric for each predicted commodity, and use this to measure how well predictions turn out in practice?
Must be some clever statistician out there who knows how to tell how well correlated odds are with outcomes, and to what extent this correlation signifies informed prediction vs fortunate confidence.
I guess roulette ends up with a very low predictability metric of 0.1, horse racing low at 0.2, share prices medium 0.3, and political events high 0.4?
You could then define gambling as the class of future events with a predictability metric of less than 0.25, say.
Already been done? Silly idea? Flawed/Meaningless?
I blogged a welcome to the new center in July at CommerceNet (http://blog.commerce.net/?p=259). The luminary to notice is Tom Malone, who has been doing work related to Prediction Markets for many years. Tomaso Poggio (who co-authored “Securities Trading of Concepts”) is also a principal. More details in that post. Now that they’ve made the formal announcement, I’m happy to welcome them again.