9 Comments
User's avatar
Matt's avatar

What are the trials we’ve seen recently that you mention?

Robin Hanson's avatar

I added links to the post there.

Carlos's avatar

I am sorry professor, but it was actually a terrible idea, because it would be a market that predicts its own result.

Imagine a market where there are two notes, A and B, and a gram of gold is paid to the winner, that is, the one that trades higher on a specific day. In that case, you have nothing but a bidding war.

The whole point of predicition markets is that they predict an external event, that which the market participants cannot manipulate.

Robin Hanson's avatar

You are describing an auction. Prediction markets aren't the same as auctions.

Carlos's avatar

How not? We are predicting whether Trump gonna attack Iran, and the actual decision to attack Iran is made by whichever prediction wins? So throwing money on either side is a weird combination of buying the decision and predicting whether that side will be better at buying the decision than the other. It has nothing to do with whether that attack is generally a good idea or not.

Nicholas Reville's avatar

Have existing prediction markets said that they won't allow markets of this type? Seems compatible with their model and washout rules, but I'm not an expert.

Robin Hanson's avatar

Manifold and Metaculus had conditionals in the last presidential election, but neither Kalshi nor Polymarket have implemented them.

Nicholas Reville's avatar

But doesn't the polymarket '50-50' resolution or the manifold 'N/A' condition allow this? Like when a sports game is cancelled, they resolve to N/A.

Robin Hanson's avatar

That's not the same as being able to create a market using a condition that you specify.