Collective Market Intelligence

At Etech yesterday morning Gary Flake of Yahoo! Labs said his organization has four research areas. I only remember three: collective intelligence, machine learning and (fairly obviously) text mining. After pointing people to Yahoo! Next, Flake launched the Tech Buzz Game. It’s a prediction market where participants bet funny money on future search traffic for keywords associated with a technology relative to keywords associated with competing technologies (e.g., the programming language market includes C, C#, C++, Java, and several others).

Either I or many game participants horribly understand how buzz scores are calculated. The game FAQ says:

The buzz score of a stock is the number of searches on any of the stock’s buzz words over the past seven days, as a percentage of all the stocks in the same market.

Yesterday Ruby was worth fifty percent more than any other language. I suspected that participants think the buzz score is a measure of relative change rather than of quantity. However, now I suspect people are voting for their favorite technologies rather than betting on results. Those players will lose on (every) Friday when prices are adjusted to reflect actual buzz score.

I wonder how weekly revaluation will impact the ability to gauge long term predictions? If participants expect a technology to become more popular over the next year (with an attendant increase in search traffic), how will the corresponding security behave week to week? Would traders consistently bidding the price of a security up and losing money at each weekly revaluation be the predictor of a long term increase in search traffic?

Though it feels toy-like, I’m gratified that this prediction market is considered a collective intelligence application. I often hear people saying that humanity needs to increase intelligence to have any hope of surviving whatever dangers are supposedly near, usually accompanied by complete ignorance of markets’ role as a distributed discovery mechanism and the potential for markets designed explicitly for information discovery.

In other idea futures news, check out open source market infrastructure to be Zocalo and its motivating proposal, to be developed at CommerceNet Labs.

Update 20050318: I was correct about scoring and revaluation. I made a 150% funny money profit after today’s revaluation, before which I had a loss. I made no trades after becoming fully invested. Will be interesting to see what happens in the next week. Will the Buzz Game merely be a “day trading” and game-rules-ignorance-arbitrage phenomenon? I suspect so. Too bad. A market structured to make predictions about technology success would be really interesting.

3 Responses

  1. jck says:

    collective intelligence ??? where ? you can pick any “market” and see things like buzz score for ms office 85 flat for weeks and this [false] market pricing it at 57 [equivalent]…it’s a safe bet that whoever designed this game is pretty clueless about “markets” what they do and how they function…there is a pretty good reason why double cda dominates in financial markets and why pari-mutuel is for the racetrack

  2. […] and the potential for markets designed explicitly for information discovery. This is from our Mike Linksvayer, circa 2005 —in a great blog post pondering on the Yahoo! Buzz Game. Read the last blog posts by Chris. F. Masse:The guy who invented LifeStraw, an instant water […]

  3. […] Collective Market Intelligence. See “Decision Markets…” above. […]

Leave a Reply