Prediction Market Problems

Chris Masse provides a lengthy analysis of an ongoing controversy about how the Tradesports prediction exchange handled its contract on North Korean missile launches.

The short description of what happened is that Tradesports defined its NK-missile-launch contract as requiring US Defense Department confirmation of “a test missile that leaves North Korean air space on/before 31st July 2006.” Thus a confirmed launch on or before July 31 would settle the contract at 100, while no confirmation by the end of July 31 would mean the contract would expire at zero. The NK government on July 4 launched missiles that left NK air space and should have resulted in the contract settling at 100. However, because the US DOD did not provide what Tradesports considers to be official confirmation that the missiles left NK air space, Tradesports has not settled the contract and many traders who were betting on the launch are upset.

Chris thinks that this dispute will discourage trader participation in, and may damage the credibility of, prediction markets. He may be right but I suspect the long-run costs will not be as big as he fears. There have been many similar instances of badly handled contracts in traditional equities and derivatives markets, and I think what generally happens is that traders note what transpired and try to avoid getting caught in similar situations in the future. At the same time exchange officials tend to quietly learn from their mistakes, as do the people who run competing exchanges.

For example, in the past few years on electronic futures exchanges there have been several well-known occasions in which traders made fat-finger mistakes where they entered buy or sell orders for, say, ten thousand contracts when they had meant to type “1”. (And where their employers or brokers had neglected to set contract-size limits in their trading software, but that is another issue.) What happens then is a huge and instantaneous price spike, followed by gradual recovery to prior levels as participants sort out what happened. Such screwups create big losers and winners, and the exchange has to decide whether to let the trades stand, or to invalidate all trades for the period defined as the mistake or to make some kind of compromise. Sometimes, particularly at exchanges that are dominated by big institutions and where the mistake is made by an institutional trader, the at-fault institution is let off the hook. Such outcomes are very aggravating, and often costly, for other traders, but trading continues.

It will be interesting to see what ultimately happens in the case of Tradesports and the NK missile contract.

UPDATE (07/23): I may have been too optimistic. Chris Masse posts critical comments from a trader in response to what I wrote. (Click on the link and scroll down.)

He also posts several comments from other traders on the general topic of TradeSports and how it defines its contracts. These comments are reasonable but do not agree on whether the contract’s terms for a missile launch have been satisfied, suggesting that TradeSports is doing something wrong in a big way. There should be no uncertainty about such a fundamental issue.

The comments also remind me of a point that I should have mentioned earlier. By handling the NK contract so confusingly, TradeSports has transformed it, for the remainder of its life, into a contract on the TradeSports decisionmaking process — albeit with a knockout provision in the event of DOD confirmation of the July 4 launch or another launch before the end of July 31. That’s not progress.

UPDATE 2: Bo Cowgill comments.

UPDATE 3: Chris Masse reminds me that Bo Cowgill is the guy at Google who is experimenting with internal prediction markets.

UPDATE 4 (07/25): Tradesports, apparently in response to complaints, introduced improved procedures for dealing with settlement controversies like the one surrounding the NK missile-test contract.

(Disclosure: This blog is a Tradesports affiliate.)

4 thoughts on “Prediction Market Problems”

  1. Hi Florida,

    You’re making a good point. Prediction exchanges such as TradeSports learn from each of their mistakes, ultimately.

    What make traders piss off is the mistake in the first place, and that the substance of their criticism is often ignored (or so they feel like it is).

  2. A market works best when it abstracts the least. In other words, the information the market deals with should be as concrete as possible. For example, commodities markets deal with information such as X amount of commodity Y delivered to location Z by such and such date. Such information is easily measured but all participants.

    The Achilles’ heel of prediction market is the relatively vague outcomes they are supposed to predict. The temptation I think is to lump many different outcomes under one umbrella ex Israel invades Lebanon. Do airstrikes count? Commando raids? etc.

    The comparatively artificial parameters of the prediction markets makes them less reliable than that for markets that deal with actual physical goods.

  3. How should the prediction markets deal with that issue? “Israel invades Lebanon” is not 100% concrete. Should the markets respond by making more specific contracts (“Israel has [NUMBER] or more ground troops within Lebanese borders at midnight on [DATE]”)?

    It seems part of the problem here is this one contract specifies something people don’t see as an essential part of the issue — confirmation from the DoD, in this case, when it’s been confirmed by other parts of the government. If prediction markets are to make their predictions more concrete, I would think they’d have to include things like this by necessity — in which case, the “short description” of the contract becomes misleading.

    Maybe this is something traders just have to pick up on — when trading something like “Israel invades Lebanon” you just have to read the contract carefully, and stay out of it if there are technicalities you worry about.

    Maybe it’s also something TradeSports and others need to address — if and when there is a grievance over the interpretation of a contract, how is that dealt with?

  4. This is interesting. To me, the untrained normal guy, Tradesports looks to be in the right. No DoD acknowledgement, no pay. Of course, there is wiggle room in just what exactly that acknowledgement is.

    The contract is written poorly, and those who wagered should have seen it.

    But what is worse? Take a look at those InTrade contracts for when/where/what category the hurricanes will hit.

    I can already see it. Is a barrier island actually part of North Carolina? When exactly did the winds die down from a cat 3 to a cat 2? How does anyone decide that?

    And what about US/Israel bomb Iran? What is a bomb? A missle? An actual bomb dropped from a plane? A rifle powered grenade? An artillery shell? One bomb? Fifty?

    I once looked at playing around with these contracts with some of my fun money, but knew that someday something like this would happen.

    Honestly, Tradesports should settle both sides of the contract at 100 immediately and then in the future either structure the contracts with ironclad language or just say that everyone is making these bets AT THEIR OWN RISK since it would be the call of Tradesports for example if that hurricane that hits North Carolina was a cat 3 or 2.

Comments are closed.