Hacker News Reader

Comment by FFRefresh | Hacker News Reader

I am a fan of systems/tools which reward those with better models of objective reality and the likely future.

I am intrigued by prediction markets as a way to give feedback to people's models of reality and how accurate they are in extrapolating trends, and to reward those who are better able to predict the future. It's a potential way to introduce 'skin in the game', both from a reputational and financial perspective.

Is that 'media expert' claiming Country X will be a dictatorship in 2 years because their political opposition won an election? Define terms and ask them to make a bet, and let's see how that plays out.

After awhile, you can start to see people's track records, and the skin in the game pressure may incentivize better epistemics before making claims.

View on HN

I don't think prediction markets make sense. The information about whether a country will become a dictatorship is usually private, and impossible to obtain. It isn't like a betting market where the relevant information is public.

I probably agree with your aims. I think that systems should exist to make better decisions but there is limit because no-one can make perfect decisions (and it is probably more important to consider robustness at that point). I think attempting to reward predictions makes no sense, everyone is prone to these biases. I don't think rewarding a certain group of people makes that any better.

One part of markets that people often get wrong is that sometimes markets are totally wrong. With Enron, the information that the company was fraudulent was publicly available, and relatively easy to find. But the price for the stock was set by people who didn't want to find it. Mayweather vs McGregor...crazy odds. In 2012, the US prediction market Intrade had wildly inaccurate odds for the Presidential election, this was a market that was trading hundreds of thousands per day. Markets work most of the time, they don't work all the time (I would say financial markets are the best evidence of this, you see stuff that makes no sense almost all the time because investors have such different aims...you think no-one is this stupid, then you see it happen in a megacap stock worth hundreds of billions, and your perspective on human reasoning changes significantly).

For the record, I don't think prediction markets are perfect and solve all problems.

I also don't think solutions/tools should be discarded because they are not perfect.

From your examples, it seems like your knock against prediction markets are the fact the aggregated predictions don't end being correct at times? Of course that's going to happen, and I don't necessary see instances of the wrong aggregated predictions of a market as a bug, but a feature. Human populations are wrong about things all the time in the aggregate. What better way to get better at decision-making than getting feedback on how we are wrong?

My argument for prediction markets was not for generating knowledge in the aggregate from all of the individual users' predictions. It was centered on the individual front as a way to reward individuals who are better able to predict the future. The uber context in which I was commenting was around who should be considered an expert. Prediction markets are one potential tool (albeit not perfect) to measure individual's conception of current reality and future reality.

Right, but the point is that prediction markets remove the subjectivity, and I am saying they don't. They are just wrong in different ways.

Rewarding individuals? So we swap one set of "experts" for another? The point of this is that no-one is an expert. The problems are created not because we don't have prediction markets, but because we treat "experts" with reverence.

The comment about randomness you made to another reply...with any event, there is some part of the outcome that is unpredictable. You could have perfect information, and still would be unable to predict it every time. Some events have no information that can be used to predict them, so predicting is just the event's previous frequency or last value.

Just for clarity, are you of the opinion that prediction markets have precisely zero value, positive or negative?

It would depend on the underlying event. The reason we have financial markets isn't to predict things, it is to fund things that need funding, and the price mechanism guides that for better or worse.

Creating an artificial prediction market is more like gambling where people are transferring risk for its own sake. Nothing wrong with that but I am not sure what the purpose would be. As said, these markets are also unlikely to add any new information either because the kind of events being talked about some largely random.

> It would depend on the underlying event.

Agreed.

Would it be fair to say that there is some value, but the quantity is unknown?

> Creating an artificial prediction market is more like gambling where people are transferring risk for its own sake.

I suppose this is part of what is involved, but "transferring risk for its own sake" doesn't seem like a comprehensive description of what they are.

> Nothing wrong with that but I am not sure what the purpose would be.

I think it's useful to always consider that there may be more to see than what catches one's eye.

> As said, these markets are also unlikely to add any new information either because the kind of events being talked about some largely random.

If I rephrase this to:

>> Because [the kind of events being talked about are largely random] it logically follows that [these markets are unlikely to add any new information].

...it seems to me that this is an imperfect way of thinking about it. Perhaps I have mischaracterized your intended meaning?

Reply on news.ycombinator.com