Google's Prediction Market

Googling Yourself writes "Employees at Google are encouraged to place bets on Google's prediction market — an exchange that tries to forecast events based on the money wagered on a particular outcome. Employees have made wagers with play money (Goobles, as in rubles) on questions like: will Google open a Russia office? will Apple release an Intel-based Mac? how many users will Gmail have at the end of the quarter? One tangible benefit to the company is that the market allows Google to track how information disseminates in the company. A paper called "Using Prediction Markets to Track Information Flows: Evidence From Google" discusses information flows in the company based on the prediction market data and contains many other interesting observations of Google culture. (pdf)"

Employee Games the system

[ryanguill]

Google, however, is no ordinary company. As detailed in a footnote, one Google employee, looking to be a profitable trader, wrote the code for an extremely prolific trading robot. As a result, he "was participating in about half of all trades" and made a profit (in Goobles).

I wonder if he got fired or got a raise?

Re:Information, not crystal ball

[UbuntuDupe]

Prediction markets reflect information that is known; they are not perfect crystal balls. They also fall victim to groupthink.

Very true, but the question is, compared to what? Prediction markets force people to be honest -- there's real money at stake, and a cost to your arrogance. (Well, not at Google, but at other markets.) It's harder to be a know-it-all when you have to back up your claims with your own money.

The best example is that the prediction markets predicted Hillary would win the Democratic nomination by a wide margin. Now the consensus is that it will be Obama by a wide margin.

I'd classify that as a feature rather than a flaw. Changes in odds are useful information too and allow us to avoid hindsight bias. It keeps us from saying, "Well everyone knew Obama would win." You can look at the history of the bets and say, "No, they didn't -- the general consensus was that he had no chance." Furthermore, "Hey, if you really knew it all along, why weren't you swiping up those underpriced Obama bets?"

One thing I'm interested in seeing is how Candidate/Party X's chance of winning correlates with critical financial metrics like long-term interest rates and oil prices. That is, do traders revise their estimates as a party's chance of winning goes up? Intrade recently started "shock future" bets where you can bet on the changes in such variables on election day (although I think to avoid the noise you should instead look at the long-term correlation rather than one day, which has noise from other factors).

For all their flaws, prediction markets truly fascinate me.

Predictions helped find a nuke, and a missing sub

[Majin Bubu]

I find prediction markets very fascinating and a powerful prediction system, when manned by experts of the field. Indeed, such a system was used to find a H-bomb that had been lost at see back in the 60s, as well as the missing submarine Scorpion. Read "The Blind Man's Bluff" for details, but there are excerpts online. Also see Bayes subjective probability.

How Prediction Truly Works:

[imstanny]

The best predictions are made by an aggregate of opinions. You need look no further than "Who wants to be a Millionaire". When they poll the audience, the audience has a very high success rate (over 90%, though I forget the precise figure). The reason for this is that people that don't know for the answer account for random choices, which allows for those that do know the answer to make the correct answer rise to the top.

There's a fascinating book that addresses all these points called "More Than You Know" by Michael J. Mauboussin. I have recently been at one of his lectures at an investment conference (and got an autographed copy of his book). He gave a poignant example about prediction quality of a group vs the individual:

He put jelly beans in a jar and made his students guess the amount of beans the jar contained. He offered a small monetery reward as incenctive, to better ensure educated guesses. With the exception of 2 students, the class average came close to guessing the amount of beans in the jar than any one individual. His book offers interesting examinations of psychology and group behavior applied to financial markets.

Re:I'm suddenly skeptical

[sholden]

You might be trying to ferret out the wisdom of crowds. Lots of other people are just gambling, and another entity is interested in collecting transaction fees...

Re:I'm suddenly skeptical

[Spy Hunter]

I don't need Intrade to tell me what's happening at the polls; I can get that from the news. [...] we're basically counting on Intrade to let the wisdom of crowds ferret out all of that before the polls

The thing you have to understand about prediction markets is they don't represent a magical source of extra information. Prediction markets work only based on information that is already publicly available. They simply combine and weigh that information better than any other method of making predictions, and that is still a valuable service.

When the markets predict John McCain winning the nomination, they are not simply reacting to recent NH poll numbers; nothing a market does is "simple". The market is weighing *all* the available evidence to come to the conclusion that John McCain has the best overall chance to win the nomination. When you look at the market's prediction, you can be reasonably sure that single source reflects all the information you would likely be able to gather yourself if you dedicated all your time to researching the elections.

The campaign contains relatively few surprises.

Well put your money where your mouth is, bub. If you know better than the market, you stand to make money on it. The truth is, the volatility of these election markets reflects the *actual* uncertainty in *everyone*'s predictions about this primary, which is further a reflection of the actual volatility of public opinion itself. There *have* been surprises. Six months ago nobody expected Giuliani to do this poorly; nobody saw Huckabee's gains among evangelicals; etc etc. To pretend otherwise is silly; if you knew these things ahead of time you would be sitting pretty with a great position in the markets right now.