Fake Prediction Markets Useless

from the it-doesn't-even-work-in-theory dept

Tue, Mar 7th 2006 12:30pm — Joseph Weisenthal

In an attempt to harness the wisdom of crowds, many so-called prediction markets have sprung up where users can bet on things like who will be the Democratic nominee in '08, or the winner of American Idol. In theory, the trading in a given contract will reveal the probability of such an event happening. A few of these markets, like Tradesports, are real money exchanges, but most of the ones that pop up, like the latest one, Inkling (it appears to have launched in the last week, and claims investment from Paul Graham), involve trading in fake money. There are several others, including one on technology trends hosted by Yahoo! and O'Reilly Publishing. But what's the point of a play-money exchange? Prediction markets are supposed to get people with information and insight to put their money where their mouth is, taking a risk for an opportunity to profit. A fake-money exchange will get participants with time to waste, who don't necessarily know much about the subject matter they're betting on. What information will this reveal? One hurdle, at this point, is government regulation which greatly restricts this type of trading. Still, with little to be gained from these endeavors, they hardly seem like a good alternative.

7 Comments

If you liked this post, you may also be interested in...

Reader Comments

Subscribe: RSS

View by: Time | Thread

Andrew, 7 Mar 2006 @ 2:24pm

Fake markets actually do work...within reason
Many market simulations fail because the experimenters fail to create a proper reward or penalty system. The markets mentioned are imprecise representations of the work of Vernon Smith (the guy that got the 2002 Noble in economics for his work in this area). Like any cargo cult approach to science, something has to actually replicate the right conditions and not just appear to.
[ link to this | view in chronology ]
John Lambert (profile), 7 Mar 2006 @ 3:22pm

There's a market for everything...
I could really use some good fake predictions. I'm always being asked how long something's going to take me when I have no basis for forecasting the time required. I'll sign up for one-a-day.
[ link to this | view in chronology ]
John, 7 Mar 2006 @ 6:16pm

Just like poker
This phenomenon can also be observed when people play poker for fake money. All of a sudden they are willing to take chances and risks that they wouldnt take with there were appropriate consequences for being wrong. A lot of people tell me they want to learn poker but then play on the fake money websites. They end up learning very little - even if they are playing a proper game with a view to accumlating as much fake money as possible, there are always enough people that dont care, such that what they learn is really just bunk and not applicable to a real money game. Same goes here - we learn nothing about trends since people arent making "real" decisions.

- John
[ link to this | view in chronology ]
?, 7 Mar 2006 @ 11:13pm

Yup
It is a time waster, a game. No money lost, nothing gained, cept a little ego if your guess prove right.

:)
[ link to this | view in chronology ]
dan mayer, 8 Mar 2006 @ 8:53am

still
still there is something to be said for the collective concious, and that is what they are harnessing. It is the same as some machine learning techniques instead of making a very complex learning algo you have a bunch of tiny simplestic learning models, like short prediction trees, then you ahve all the very simplestic models vote. You set some sort of voting threashold and viola you have a very high success rate...

so get a bunch of pretty dumb (about the subject) people with limited knowledge and it might work the same and predict some very interesting results.
[ link to this | view in chronology ]
CrowdIQ Team, 13 Mar 2006 @ 8:03am

Love to get your thoughts...
...on our little project. We're almost identical to Inkling, minus, the UI and Trading Engine Abstraction. But we're in the process of building our trading community and hope we can become the eBay of prediction market.s
[ link to this | view in chronology ]
David Pennock, 29 May 2006 @ 4:23am

Fake money sometimes good enough
I agree with some of this reasoning (e.g., the poker example) but, perhaps surprisingly, research shows that in some circumstances, fake-money markets are just as accurate as real-money markets in terms of their predictive ability:
Prediction Markets: Does Money Matter? - (PDF) , Electronic Markets, 2004
The Real Power of Artificial Markets, Science, 2001
[ link to this | view in chronology ]