The "gaming" of anything is generally bad. In the case of prediction markets, there are occasionally some which easily be gamed by a broad audience. There is a new market on hubdub for:
How many results will a Google search for 'Switchy McSwitchenstein' yield by the end of April 2008?
There are some problems with this question from the beginning (Google searches vary based on a lot of different settings and also based on which data center you get). One other problem is that the market itself provides incentive for individuals to "game" with the market itself. It's quite easy to do - this page will be part of the Google index shortly. And this. See my point?
The hubdub "good questions" faq doesn't really cover this issue.
My feeling is that most prediction markets are around to help predict the outcome of some event. They should be designed to limit the number of people who can "game" the outcome. Otherwise market participants will play with the outcome in order to win a bet and the value becomes meaningless.
My advice to market admins: prohibit these types questions in your guidelines and review them early to mark them void.
Also, this post is the beginning of a new series of posts on the topic of policies for effective markets.
Update: The folks at hubdub voided the question with the following message.
This question was voided. Reason: Too easy to game; results differ by user; Google search questions now banned (Rule 2.2.5)
So, I guess that answers how they feel about it.