An Introduction to Prediction Markets

Nitpicky correction: "Metacalculus" is actually called "Metaculus". Its name is not derived from "calculus".

Whoa, I always read it as Metacalculus. Editing on both thanks for the heads up.

Let’s say you are extremely confident that the Tokyo Olympics will take place this summer. You could place a wager on the Yes outcome by buying Yes shares at $0.91 per share (Total cost, ~$0.93 per share, as there is a 2% Liquidity Provider fee for market makers).

Previous suggestion on LessWrong suggests that savy uses of PolyMarket don't do that. They would create a Yes/No share pair and then sell the No for a few of 2% of $0.09.

While arriving at an accurate probability that a sporting event will take place is perhaps not as useful to society, when it comes to markets in health, economics, and politics, there is utility to forecasting an accurate probability of an outcome.

I don't think that's the case. Having access to an accurate probability about whether the Olympics will tell local hotels about how important it is to have a lot of beds available. It will tell AirBnB whether it makes sense to run ad campaings to get people to rent out their homes for hosting tourists that come for the Olympics.

[-]Annapurna5y10

Previous suggestion on LessWrong suggests that savy uses of PolyMarket don't do that. They would create a Yes/No share pair and then sell the No for a few of 2% of $0.09.

This is true. Polymarket has a couple of tricks to save on fees.

I don't think that's the case. Having access to an accurate probability about whether the Olympics will tell local hotels about how important it is to have a lot of beds available. It will tell AirBnB whether it makes sense to run ad campaings to get people to rent out their homes for hosting tourists that come for the Olympics.

I agree with you. Didn't really think of this to be honest!

[-]Bojadła5y10

Previous suggestion on LessWrong suggests that savy uses of PolyMarket don't do that. They would create a Yes/No share pair and then sell the No for a few of 2% of $0.09.

Could you explain this more or link to a previous suggestion? I don't get it.

[-]ChristianKl5y30

https://www.lesswrong.com/posts/nezKGutEuwYGhK29j/exploiting-crypto-prediction-markets-for-fun-and-profit is a previous post that explains how to interact with prediction markets and it's in the comment section of it.

[-]SimonM5y10

Having access to an accurate probability about whether the Olympics will tell local hotels about how important it is to have a lot of beds available

Aside from changing pricing on the rooms (which is already an implicit prediction market on the Olympics) I'm not really sure what the hotels are supposed to do. Individual hotels can't exactly increase supply overnight. (Unlikely your example with Airbnb)

[-]ChristianKl5y20

If you set a price for a room months in advance knowing the likely demand for the room at the time it's used is very useful.

Generally, running the Olympics comes with a lot of local economic activity to make the event happen and various actors benefit from being able to plan ahead.

[-]SimonM5y10

Generally, running the Olympics comes with a lot of local economic activity to make the event happen and various actors benefit from being able to plan ahead.

I agree with this, I just don't think hotel rooms are a particularly good example since supply is fixed there is little hotel operators can do with knowledge of the probability of the event. (They can "change" prices, but prices are effectively driven my a market equilibrium (which is going to effectively be a prediction market on the Olympics going ahead))

[-]ChristianKl5y20

The bed market is similar to a prediction market but less efficient then a good prediction market. If people buy hotel beds and then resell them that adds a lot of overhead that you don't have if that market function is done through a good prediction market.

[-]SimonM5y10

To give one further example of uses of prediction markets, Scott Sumner has the idea of using NGDP futures as a way to have market driven monetary policy.

That said, whenever these things become more "useful" I can't help but worry that "information discovery" and "institutional hedging" act against each other. Ultimately if something becomes correlated to people's view of the world, pricing becomes less about "forecasting" and more about "risk premium". To take a concrete example, if there were NGDP futures, you should expect them to be biased low / you should earn a premium for being long GDP. Specifically because if GDP falls so will the rest of your assets, which makes it "painful" to hold - hence a premium to own it.