Fixed; thanks.

Expand full comment

There have been many dozens of internal prediction markets that have worked fine. With a market maker such markets can work fine with only a handful of traders. Even when individual trades are anonymous, the overall trader score need not be anonymous.

Expand full comment

Wouldn't an internal (within-firm) prediction market be too thin to work? Also, without anonymous trading, what incentive would employees have to reveal information by trading on such markets?

Expand full comment

"It matters if you want to know whether industry and government should invest resources in prediction markets or in predictive expertise (represented, in this "experiment" by pollsters)."

You mean if there are only enough resources to conduct polls, or have a prediction market (or have a half-ass poll and a half-ass prediction market)? In that case, yes, you would definitely want to know what prediction markets can do without using poll data. I can see how such resource limitations could actually be a problem for certain other topics. It would definitely be useful to answer the question of spending the resources you would spend on an Obamacare website prediction market instead of spending that money on a commission of independent experts would yield better predictions or not. After all the prediction market does need some inside information, otherwise they predict the discovery of Saddam's nukes again.

Expand full comment

It matters if you want to know whether industry and government should invest resources in prediction markets or in predictive expertise (represented, in this "experiment" by pollsters). (The electoral predictions are the most visible arena where prediction markets are said to have succeeded.)

Robin says a prediction market would have given Obama advance notice of his problems with the healthcare site. Perhaps, but would not an administration less hostile to whistleblowers have gotten better technical information to the President? (In other words, expert opinion [analogous to pollsters] is preferable to the output from prediction markets.)

In the end, it depends on their relative expensiveness. But the proponents of prediction markets seem to deploy a double standard. When faced by attacks on the efficiency of markets, they reply by saying that relative efficiency is the relevant consideration. But when it comes to their own evidence for prediction markets, they are uninterested in (unbiased) tests against the alternative approaches to prediction.

Even if prediction markets would represent an improvement over today's corporate decisionmaking, their achievements owe to overcoming the yes-man factor. And if industry were interested in overcoming that, there would (seemingly) be more efficient ways than using prediction markets. At least we don't see the proponents here weighing prediction markets against less radical counters to the yes-man problem.

Expand full comment

What? Hedge funds are famous for using extremely complex trading strategies which rely on finely-balanced probabilities and deploying huge sums to pick up tiny advantages

Yeah, that's why hedge funds were moving so much money on Intrade. Oh wait, they weren't.

and this is most dramatically apparent in instances like LTCM

Uh? According to Wikipedia that was a hedge fund that accumulated an enormous debt, failed, and was bailed out by the government. If that was your best example of market efficiency then you are not really helping your case.

Expand full comment

"Given the conditionality, attempts to fight the manipulation aren't much disadvantaged - if the condition falls through, no money changes hands, after all. What's sauce for the goose is sauce for the gander..."

Some markets will have conditionality, others won't, in a conditional market the anti-manipulators do indeed have less of a disadvantage, but the manipulators won't lose money if their manipulation fails, so it's just neutral to them.

"Doesn't make the slightest difference. Do you think the Kerry and Romney manipulations came with signed cheques? Nevertheless, the Intrade forums and blogs erupted with accusations of manipulation."

Let me guess, two camps accused each other roughly equally and there was no way for people to know whether those were baseless accusations (no one was manipulating), half true (only one side was manipulating), true (both sides were manipulating or three quarters true (both sides were manipulating but the relative magnitudes were quite different from those of the accusations). Good luck coordinating an anti-manipulation campaign among all the noise. Disinformation is a b*tch.

"The existing levels of manipulation like LIBOR hasn't stopped the financial markets from being overall pretty efficient, either."

First of all the financial market is always right even when it's not right (because it sets prices as well as trying to predict them: when the financial market assesses houses as more valuable than they really are then housing prices still go up), second, the financial market is very broad with many powerful competing interests, prediction markets will be much narrower and smaller (combined they may be very broad and large, but each individual market won't be).

"As your own examples point out, people do not need to lock up arbitrarily large sums for long periods on their own account: if that's really an issue, they can invest in mutual-fund-like corporations"

Good point, but then they'll be left to the investment choices of those funds. In any case I think you really underestimate economic inequality: the richest 1% of the world own ~50% of the wealth, more if you look at liquid wealth and as I pointed out governments can also play the game.

Expand full comment

"You know, *actual* weather forecasters uses large ensembles of models with human input to gain predictive edges (I think Silver estimates the human addition as ~10% increase in accuracy over the current state-of-the-art models) - but oddly enough, they think this is a good thing and not a bad thing."

There are only a few models. Meteorologists the world over pretty much agree within strict boundaries which combination of models to use for a certain timespan for a certain location. Investors will go with this advice and, like meteorologists, will not disagree over whether a typhoon will hit China the day after tomorrow or not.

Expand full comment

Criticism based on a single pair of contracts on a single day where it's not obvious the movements were even wrong. This doesn't even begin to rise to level of a statistical demonstration that prediction markets are inefficient up until Election Day - which was your claim!

Epic fail.

Expand full comment

Wow, I'm impressed that you managed to read through all 4000 results and dismiss the Intrade, IEM, HSX, Google, Betfair etc results are being "no real-world data". Well, I will just have to defer to you on this matter since you're clearly far more familiar with the topic than me.

Expand full comment

> What cherry picking? Were there many other statisticians with a track record similar to that of Silver and Wang who tried to predict the election and performed worse than Intrade?

Even in 2012, Intrade was not the worst normal predictor (Margin of Error did worse with some regular models), and as I've pointed out ad nauseam now, the Intrade 2012 results were predictable in advance to be inferior solely on the fact that American bettors - like myself - could not trade on the markets.

Expand full comment

> Yes, and that explains why election prediction markets only become accurate near, or even on, election day

Cite or GTFO.

Expand full comment

> Hedge funds won't risk betting a large sum on the success of a policy unless they are very much confident of its success.

What? Hedge funds are famous for using extremely complex trading strategies which rely on finely-balanced probabilities and deploying huge sums to pick up tiny advantages, and this is most dramatically apparent in instances like LTCM. Where on earth is this claim coming from? Where is your evidence for this extraordinary claim?

> The accuracy of any predictor is boundend by the fact that investing on increasing it runs in diminishing returns past a certain point.

Which is why we have leverage and hedge funds can be levered by dozens of times - to amplify even small predictive advantages into huge profits.

> Therefore relatively safe manipulation is certainly possible

Again, it's no more 'relatively safe' than shorting or selling insurance are: you may make small profits in the interim, but if you ever fail, you can lose everything.

Expand full comment

"This is no unbiased contest: the investors/gamblers were free to use the results of the polls!"

Yes, and that explains why election prediction markets only become accurate near, or even on, election day. On the other hand, why should it matter what sources prediction markets use if you just want information about the future (at least as long as no one is stupid enough to reason that since prediction markets work polls can be eliminated)?

Expand full comment

> all-pay auction

Actually, it's all-losers-pay, which means that if you know you are the biggest fish, you have an incentive to go for it.

Expand full comment