Good clarifying questions - I've been in this world so long I forget
what it looks like to a normal person :).

The types of online games where I think this could work would have to
be games where there was some sort of in-game currency. MMORPGs all
have this, but some other games do as well.

For instance, my company (www.allinplay.com), has a Texas Hold'em
game. The chips played with are fake, but they are persistent-
meaning that what you win you keep form game to game. Therefore my
players prize their chips and don't bet them as recklessly as one
might think. This is also true of MMORPGs where you have to work to
earn the in-game currency and losses of it are a pain.

You bring up a great point that I had not thought of - people may vote
for the most expensive features!

I THINK such a problem can be addressed by insisting on a metric that
centers on profit or profit margins. This would allow players to
propose feature ideas and then the programmers could weigh in with
cost estimates. So at first people love the "get rid of lag" feature,
but then when the programmers point out it would cost $1 billion and
drive the company to bankruptcy... people will not vote for it.

Yet, if people suggest something that the dev team THINKS will be
expensive, and then there is a dialog that shows it need not be THAT
expensive... that could make life interesting!

If profitability is a little too cold (after all, no one likes
companies to be profitable :)), perhaps more PR-friendly metrics could
be used. Perhaps a combination of "pays for itself" (people usually
think "paying for itself" is fair) and usage (# of man-hours people
spend playing the game) or population (# of customers)?

I imagine futarchy should not be step 1 in the experiment. I think a
game company should look at what HP has done with its prediction
market and copy that first. That will give them a feel for how such a
tool could be powerful.

I imagine that if an MMORPG dev team embraced prediction markets, they
would follow a process like this:

1. Internally draw up their list of potential new features.
2. Assign creation cost estimates to each.
3. Segregate the list based on cost. So if there were 20 feature
ideas, break them up into an "expensive" list of 10 roughly-equally-
expensive ideas and 10 "cheap" roughly-equally-expensive ideas.
4. Create a prediction market for the "cheap" list, and use a metric
like # of customers.

If this works, then they can start running experiments on more
expensive features.

If that works, they can experiment to find what metrics would allow
them to pit low impact+cheap vs high impact+expensive.

Once they have prediction market code up and running, they can ease
their way into futarchy by allowing certain small (and cheap) aspects
of the game to be "ruled" by futarchy. For instance, instead of game
features, maybe people could propose community development ideas
(something that only costs the customer support staff time, not code).