Last night I went to the UUASC meeting and saw a really interesting presentation on Yahoo! Research. One of the projects that was highlighted was a project called The Buzz Game. Users get fake money to buy shares of words that might become popular. If they become popular, the price goes up and the user makes money. The person with the most at the end wins a prize. This got me thinking about prediction markets in general.

I think an interesting source for data would be using blogs. (which seems kinda obvious to me, but I couldn't find anything with a quick search) The game could work like this, you submit your blog to the website. By posting on particular subjects, you are effectively buying their shares. To even out frequent and infrequent posters you give a fixed number of 'points' per day, which are divided by the number of subjects you post on. If you are the first person to post on a particular subject then you are buying low, the last person, you buy high. At the end of the day, the blogs with the most points are the most accurate at predicting trends.

You could then come up with an interesting Cluster Graph of which sites predict which terms the best. I bet Slashdot is good for 'linux' but bad at 'fashion'. But, who knows, geek chic could be in.


posted Aug 5, 2005 | permanent link