DumbAgent.com

We at Dumb Agent have been long time fans of Predictify, not only as a source of fun bets, but as a plausible indicator of future trends, much like the Iowa Electronics Market started to be years ago, and other trading platforms such as Betfair and Intrade have become. We liked Predictify particularly, however, because it was not a gambling site per se. It is free to predict (or bet) on anything you’d like. Most predictions will not yield any money, but some will. The money for these, however, is put up by the person asking the question. This means that if you would like the Wisdom of crowds to help with a certain question, and would like the maximum response rate, you can stake your own money into it. The people predict for the money, and you get the results you were looking for. The Washington Post, New York Times, and San Francisco Chronicle all have apparently used Predictify to judge reader interest in stories.

We also like the fact that anyone can ask a question which then anyone can respond. Predictify has been following this model for several years now, so we asked them if they had any raw data or statistics regarding the accuracy. We have been rebuffed in this regard, but we suspect they do (their own site advertises their accuracy in political questions). If anyone has a method of obtaining these results, or if any intern would like do scan through them for extra credit, we would love to see some of this data. For example, is there a minimum number of predictors in order to achieve acceptable levels of accuracy in predictions? Are questions that offer cash incentives really always more precise? Are predictors more accurate in sports than politics? In Current events than pop culture? Are predictors from certain demographics more accurate than others?

These are all titillating questions which Predictify has hitherto refused to supply us an answer with. We will keep trying, however, and let you know as soon as we are successful.