This question is an attempt, similar to https://manifold.markets/dreev/biden-quiescence and https://manifold.markets/jack/will-biden-be-president-on-915-reso and so forth, to find a self-resolving mechanism that works well.
At close, I will roll a twenty-sided die. On a roll of 20, I will resolve this according to my best judgement. Otherwise, I will resolve this to the market probability.
The idea is that this sort of market is mostly, but not completely, self resolving. The creator needs to invest the time and effort in judging the resolution of only 5% of the markets they create this way, and similarly the market participants only need to actively trust the judgement of the market creator in 5% of the cases.
So while this mechanism doesn't give us a true self-resolving market, at least it gives us a 20x force multiplier for resolution capacity. An interesting further question might be how far this could be extended: if we successfully create a 20x force multiplier this way, could 100x also work? What about 1,000x, or 10,000x?
1,000
3.00I have a great randomization technique for @zzq here:
@Yev Score another one for team "identify and showcase vulnerable market mechanisms so that people don't rely on them for something that's actually important"
@jack I'm honestly baffled as to why people are spending so much time trying to find convoluted ways to have "self-resolving" markets.
@MattP Often mostly for fun I think. But as I laid out in the description of my market, there is clearly demand for market mechanisms to ask highly subjective questions: examples include:
It's relatively easy to ask "Will proof emerge" or "What will be the expert consensus" - but it's a completely different and much more interesting question to ask "What is the market's credence that this is true based on all available evidence"
And I think it's valuable to find tweaks to make such markets work better, and valuable to have experiments demonstrating how they can go wrong.
it's a completely different and much more interesting question to ask "What is the market's credence that this is true based on all available evidence"
But that is not a question that any of the self-resolving markets ask, and I'm not sure if that's a question you can ask.
If you have some probability resolving to "Will proof emerge", then that's what the market will predict (assuming it doesn't fail).
If you use a poll, that's not much different from "What will be the expert consensus", except you're polling non-experts instead of experts.
But that is not a question that any of the self-resolving markets ask, and I'm not sure if that's a question you can ask.
A self-resolving market can attempt to ask whatever it wants, of course it might not actually give the correct answer. Other examples include "Should Universal Basic Income exist in the US?" or "Is betting to the correct answer right before resolution bad?"
If you use a poll, that's not much different from "What will be the expert consensus", except you're polling non-experts instead of experts.
When I say poll, it could be a poll of anybody, or a poll of experts, or whatever makes sense in a given context. And it's slightly different because there might not be an expert consensus, and experts in many fields are bad at probabilities.
Do you plan to do the dice roll in a way that can be publicly verified? I notice you have a lot of yes shares so while I think you're probably trustworthy, it might help to reduce possible questions of election integrity. There are many possible ways to do it, such as using blockchain hashes or random.org or perhaps streaming the random roll. (I can go into more detail about it after close if you want to use one of these methods.)
I put together a quick post summarizing some of the methods I've experimented with in the past: https://manifold.markets/post/public-randomness-sources
@jack Indeed, I'd like to roll the die in a verifiable way, though I didn't think through up front exactly how I'd do it. Those ideas you propose in your post are good; let's use the first of them, https://avkg.com/en/daily-random/.
How about this: Of the daily random numbers for tomorrow, November 2nd, I'll take the third of the three random numbers in the 1-100 range, modulo twenty. That is, if it's 20, 40, 60, 80, or 100, this market will resolve YES, otherwise it will resolve to market (5%).
An example for clarity: If we were running this randomization using today's numbers, the third of the 1-100 numbers is 97, which we'd treat as a die-roll of 17, and the market would resolve 5%.
Does this make sense to everyone?
@IsaacKing Yes.
A better way to structure markets like this is to have a 95% chance of resolving to N/A, and a 5% chance of resolving to the answer to the question.
Here's an example of that method in action: https://manifold.markets/jack/does-gdpr-require-selfserviceautoma.
This does provide the 20x "force multiplier" for resolution capacity that was mentioned above. However, it also means that there's 1/20th as much profit incentive for making predictions/trades.
@IsaacKing There is a pattern of limit orders that can force the market to resolve at 99% or else give the limit trader profit but it requires a million mana. Suitable for @NinthCause only.
@IsaacKing The core idea was to be a symmetry-breaker. The problem with just always resolving to market is that the subject of the question is irrelevant to its resolution. There is nothing anchoring the market value in reality. Those pushing the price up have no advantage or disadvantage over those pushing it down.
By having a small chance of resolving to the correct answer, the symmetry is broken. One side has an edge over the other: it's not just manipulation vs manipulation. Someone who sets out to manipulate the market up has an edge over one who tries to manipulate it down, and who wants to bet against someone who has an edge over you?
It didn't actually work that way this time, but I'm cautiously hopeful for @Yev's variant of this, https://manifold.markets/Yev/will-biden-be-president-on-october-fb3d01633429, which combines this mechanism with a quiescence mechanism. Another experiment I might run is to combine this mechanism with a randomized resolution time.
@zzq Yeah, I think there is some promise here, but it also potentially relies on well-funded traders putting in a lot of mana and taking on substantial risk. So I think it's definitely better than pure resolve-to-mkt but clearly not that robust. Will be interesting to see the results of the further experiments.
I had another idea which I intend to run at some point: The probability of normal author-resolution increases as the volume of M$ increases (perhaps on a log scale). There are a couple motivations for this - one is that it should allow authors to focus their resolution capacity on the most traded markets. And this would tend to counteract the issue where amount of manipulation tends to increase as trading volume and therefore potential profits increase. Also, in theory one could charge fees on the trading and use those to fund resolutions that are expensive to carry out.