Will I be able to create matrix markets on Manifold by end of 2023?
Basic
65
12k
resolved Jan 1
Resolved
NO

With binary markets, the probabilities can change meaningfully depending on the time period and the exact criteria threshold that is being measured.

This results in 1 of 2 problems:

  • (i) You miss out on insight by limiting the question to what you are asking. E.g., by asking in this market whether this feature will be available by 2023, I'm missing out on how the probability might differ by 2024, or 6/30/23 (which could be meaningfully different).

  • (ii) You spam traders with many binary markets with slightly changed resolution criteria (e.g., one version of this market closing in 2023, one in 2024, one in 2025, etc.). I've done this before - it's not a great experience for either the market creator or the trader.

Hence "Matrix Markets". Basically, you can create a set of markets that are sensitized by up to 2 variables, but are otherwise the same for every question.

An example of a Matrix Market could be the question of what % of uninsured deposits will be returned to SVB debitors. That question could be sensitized across the variable: "% of uninsured deposits" - i.e., 50%, 60%, 70%, 80%, etc. And it could also be sensitized across time: end of 2023, end of 2024, next 5 years, next 10 years, etc.

This question would become more interesting with a "Matrix Market", where the probabilities from the submarkets are graphed and you can click into each submarket to bet at different sensitivities.

I don't think this spams markets. You would likely still have to pay the same amount per new market to maintain initial liquidity thresholds, so market creators would still be incentivized to limit sensitivities to only the most important thresholds. If anything, this results in less market spam, as creators like myself want market probabilities that are sensitized across variables without having to spam our followers and having a bunch of manual duplication work on the creation side of things.

I may trade on this question.

Get Ṁ1,000 play money

🏅 Top traders

#NameTotal profit
1Ṁ515
2Ṁ228
3Ṁ93
4Ṁ87
5Ṁ85
Sort by:

I was just thinking about this yesterday. Could just do dropdown lists for different variables and have the default interface be mostly a 1d multi-choice market.

Instead of matrix, you could also do >=2 dimensions if you don't add any bespoke UI/AMM for 2 dimensions

~~~
How satisfied will I be EOY [2024, 2025, 2026]?

Options: [Choice of next project to work on in 2024] [Hobby I take up in 2024]

- Satisfaction rating 0-2
- Satisfaction rating 2-4
...
~~~

This is a bad example because project/hobby don't factor interestingly/this seems like it should be 2 matrix markets. I wonder if there are good examples. You could also imagine allowing the 1-d multi-choices to be a function of the options selected in which case probably the user should use the API or there could be some template thing when creating the market.

Problem ii is solved by the new independent multi choice markets, but I'm not sure whether they count as matrix markets.

I have made a 3d matrix market as proof of concept.

/MartinRandall/trump-vs-biden-2024

@CarsonGale thoughts?

So this would differ from a multi-asset cpmm or multiple cpmm market because you could be asking for essentially a timescaled cumulative probability distribution function as opposed to a density probability distribution function. Or would you consider a multi-asset cpmm market which could return a density probability over timescales buckets satisfactory to resolve this as yes?

Related:

3D markets by 2024!

I hope so.

Cool idea!


We have a proposal for numeric markets using a new multiple choice mechanism which could function over one variable.

It should be possible to extend it over two variables. The underlying binary markets would track if the result lands within a particular combination, e.g. 10-20% uninsured deposits and Year 2. Then you could have an AMM on top that lets you buy any combination, including ranges like >10% and less than 3 Years.

I love that idea. I think a corresponding scatter plot graph showing the probabilities of the underlying sensitized markets would be helpful for interpretation.

The accompanying AMM seems helpful and super cool, but almost ancillary to the core innovation that doesn't require anything other than an interface modification to show sensitized binary markets alongside each other and bulk market creation across whichever variables you select.