Manifold is awarding $30,000 over the next three months to projects designed to benefit the Manifold community. (The first payout date is December 15th, and that's the one this market will resolve based on.) I submitted my custom search function and trading dashboard for a grant under this project.
A valid answer to this market is a feature suggestion or bug report. Any answer that I do not implement by December 15th resolves N/A. If my project receives at least $500 as part of the December 15th payout, all answers that I implemented resolve YES. Otherwise, they resolve NO.
To rephrase, each answer implements the question "conditional on [feature] being implemented, will Isaac's trading dashboard get at least $500 in the first payout?" (Determining causality is hard, so it's irrelevant whether the feature caused Manifold to award the money.)
That means you should bet YES on features that you think make Manifold more likely to see this project as benefitting the community, and bet NO on answers that you think they won't care about. (Given that I have limited development time and any time spent on one feature trades off against another, you actually want to bet on "how useful is this feature relative to how hard it would be to add?".)
Any answer that is a duplicate of an existing answer, or that I had already implemented before submission, will resolve N/A. Feel free to submit variants on existing answers if you think the difference is relevant.
For feature ideas, see /IsaacKing/what-changes-will-i-make-to-my-mani
What does this checkbox do? I have clicked "Add" on a few markets that I would like to see in a dashboard, but checking/unchecking the dashboard checkboxes doesn't seem to show or hide these or others.
@BoltonBailey Good suggestion, I'll do that. Your dashboard is just a custom list of markets you can view. Looks like it's broken ATM, sorry about that.
We're on Github! Feel free to submit new features and/or point out just how terribly the current ones are implemented.
https://github.com/KingSupernova31/manifold-search
@TheBayesian Yeah this market structure is a little weird, I wanted to see how it would go. I think the incentives are still correct though? It's positive EV to bet to your credence on each option individually, taking into account for Kelly considerations that all options that resolve are going to resolve in the same direction.
I'm open to alternative market structures for next month. All I want is some way to have option probabilities reflect their usefulness, so that I have an invisible-hand-approved ordering on which features I should prioratize.
@IsaacKing The incentives are correct but much weaker, I think. Like if any one option won’t ever change the probabilities by over 10%, that means nobody can make over 10% returns in expectation by betting on the better options? Which is tiny considering opportunity costs etc. But it certainly works, probably is unintuitive to ppl and imo an option being at 27% and one being at 79% is unrealistic for example, and I can’t think rn of a strictly better structure.
@TheBayesian I think strongly differing percentages are fine? As a simple example, the answer of "make the site do literally everything that anyone would want it to do, using this library I provide for free that takes you 5 minutes to implement" should be at ~100%, and the answer of "change this completely irrelevant thing that will take you weeks of work and prevent you from having any time to work on the other options" should be at the base rate of 30% or whatever.
But wait, I've realized a problem; people have to update on what other options I'm likely to implement and what they'll cause. If that 100% option from above were to be added as an answer, then obviously I'll implement it, which makes the probability that the project gets $500 nearly 100%, which means that all the other options can be bet up that high too.
So this market structure works in a vacuum, but as soon as my decisions of what to implement are based on the probabilities here, it breaks horribly.
Hey @jack, you're really good at thinking about market incentives; any ideas on how to accomplish what I want here?
I think the core problem is the strong correlation, so we need to break that. Maybe something like "which answers, if implemented, will lead Austin to cite that specific feature as a positive?" But then they should all be near 0...
earlier for example, I saw stuff get funded on manifund which updated me on the likelihood that you get funding, and so I bought up every option a bunch, and per question I could probably make 30% profit in expectation (unless I’m misjudging the situation), whereas I didn’t bother making sure the bets were in the same order they previously were in, bc the alpha in deciding that one is a few % higher than another is tiny. The question of whether an option will resolve YES/NO or NA, bc NA increases the opportunity cost, is a higher concern on my mind if anything. So yeah the incentives are right but the incentive for ‘find % chance of getting funding, and buy til that % for the options that are likely to be implemented’ takes up such a big part of the market price for each option that differentiating the relative value of each feature is tiny (but maybe this is rude to say and I don’t mean to be rude, so I will still reread the options and move them a bit to still be informative for the purpose of your decisions)
Tagging @Ernie and @RobertCousineau as you've been two of the most interested people. I would love to get your input on what I should focus on!
@IsaacKing graphic improvement for scanning, plus sparklines plus clear tools to compare between markets
Ability to see which markets I hold already and how much