- Prediction markets can be used to predict the outcomes of prizes
- This allows us to rank items by the chance they'll win
- In this case it's less valuable because the outcome is only a month away, but this system could be used to rank entries to bigger prizes, funding applications or forum content
- An example market
- The current ranking of Open Philanthrop Cause Prioritisation Contest articles according to actual prediction markets
Good rankings are valuable. They let us do the most valuable things first. If we had a well-trusted and accurate system of ranking interventions, then much of EA's current work would be trivial (and so we'd have to move to more neglected work).
One way to rank things is to get someone you trust (eg 80,000 hours), to rank a set of objects (eg causes). Now you have an ordering that you trust. But what do you do if they haven't done their ordering yet?
Well, you can predict it.
You can get people, incentivised by profit, to predict the final ranking. Markets encourage the sharing of information - people want to make money. If someone has information that an object will be highly ranked, they buy shares in that market, and will eventually make profit. In exchange, they share their information with everyone else. I think we should probably have more summaries of economics (and feminism) in EA, and if we did, we'd know this is uncontroversial among economists.
In this test case, Open Philanthropy are running a Cause Prioritisation Contest. They are looking for new causes and will award prizes to those that win.
Manifold Markets have created markets for each of those entries as to whether they will win the competition. Anyone can participate and win fake money (which can be turned into real donations to givewell) as well as a cash prize. This provides great incentives for people to predict correctly
The results are in a month, why should I care?
This ranking will suggest to you which articles are worth reading and understanding. But yes, it's not that valuable here. But it serves as a prototype for when you might care a lot more about the final results. Many things need to be ranked and will be in the future. This is a good test run.
- Here is an example market. Once you log in for free you can bet on it
Top ranked entries and their markets, in order of implied % chance of placing 1st or 2nd (13:10 UTC+1 22/9/6)
(Link to all markets homepage)
If they are in the wrong order, it's because I have to adjust them manually or because I did this quickly. If you'd like to help me in this, DM me and I'll make you a co-author. Here I argue for a post type which anyone can edit (community posts) and here that the forum could pulled in this ranking from the manifold API to rank things like this (new ranking systems)
As elsewhere the fake money can be given to GiveWell as real money, which seems enough for me to call it "actual bets"
Or places 2nd
Seems that sorting by top posts somewhat tracks with the highest % forecasted (right now the top 3 top posts are the highest %).
I would appreciate anyone who downvoted this explaining why. I think this is a straightforwardly information giving post.
Sorting by 'Highest %'age on the manifold group, I get a different list -- it seems like you might have missed some?
Entirely possible, though in my defence I think they added that sort because I asked them to. I will as later, let me know if you want to become a coauthor and add.
Things I will pay/be grateful for someone to do: