Should markets auto-resolve-to-PROB if the creator doesn't resolve them?
This is a proposal that resolves to whatever the consensus is. I'm betting on my ability to make a convincing case! If the founders voice agreement/disagreement, it'll resolve YES/NO. Otherwise it resolves-to-PROB, beauty-contest style, post-quiescence. (That means whatever the final market probability is, that's what this resolves to, but the end date is extended until trading has quiesced.) Problem: It's super annoying when a market is closed but the creator hasn't gotten around to resolving it. Everyone's money is tied up, and who knows if the creator will ever actually come back. Sure, the creator earns a fee for resolving a market but I contend that no dangled carrots like that are going to put much of a dent in the problem of creators wandering off. Related problem: The "closed" vs "resolved" distinction is confusing and slightly messy, having to remember to specify "by X date" separately from choosing a market close date. Proposal: Have a single creator-chosen market deadline when the market auto-resolves-to-PROB if the creator doesn't resolve it before then. The creator is still free to change that date at will. If they want to keep extending it indefinitely, they can. Just that if they ever wander off, everyone knows when the market will resolve. It's similar to the current market close date, just that there's no such thing as resolving *after* that date. And to prevent sniping (pushing the probability to something crazy right before the deadline) also have the standard quiescence rule: the market doesn't close until the deadline is past *and* an hour has gone by with no trading. I think only a genuine deadline -- i.e., with genuine consequences -- will induce creators to getting around to resolving their markets. I also think that the final market probability is the best we can ever do for resolving a market without getting a verdict from the creator (that's the whole beauty of prediction markets, in fact). Related markets: 1. https://manifold.markets/MattPrice/will-manifold-implement-either-repo 2. https://manifold.markets/M/what-happens-when-market-creator-le-a60924d5a2b4 3. https://manifold.markets/IsaacKing/will-manifold-allow-us-to-set-a-res 4. https://manifold.markets/Undox/market-resolution-is-yes-but-undox In conclusion, I propose that every market have a single deadline that serves as both market close and latest possible market resolution, automatically resolving to the market's own consensus if the creator doesn't resolve it. And *this* market resolves to how good an idea that is, as determined by market consensus (or the founders weighing in).
Maybe the creator could choose the default resolution? I can see other use cases for this - if you have a market like "Will aliens invade Earth by 2023?", it could be handy to have that resolve NO by default at the end of the year (plus a delay). That is an extra setting, though. Seems like the default for that setting should be N/A, even if PROB is better in some circumstances.
If MM decides they *must* have some standard policy for abandoned markets, the minimum delay required before it's automatically resolved should be equal to the delay from market creation to market close. There must be a very long lag between close time and auto-resolve time.
I wonder if there comes a point where a market gets so big that admins should take over the resolution anyway - take away any risk of rugpulling or misclicking (and make the creator's job a teeny bit easier). (Given that the question is objective enough, anyway.)
Crappy proposal that shouldn't be adopted but might give food for thought: what if every market, in the period between closing and resolution, had a meta-market predicting what the resolution would be, conditional on actually being resolved? Then resolve to whatever probability that market gets to (if the creator doesn't actually resolve). If the true resolution is publicly known, that market will converge to it pretty quickly. Otherwise, it'll probably converge to the original prediction anyway. (Of course, if the default resolution is N/A, every market becomes conditional on actually being resolved...)
@dreev, that was pretty much what I was thinking, yeah. I think that makes sense even if the default ends up being to auto-resolve N/A instead of PROB (which I'm not fully decided on either - N/A seems fairer in a way, but PROB at least maintains some truth-correlation). I'd argue for a week as the default delay. An hour's way too short, a day might be okay but I'd rather be able to have a weekly check-in be enough to make sure all my markets get resolved properly.
Auto resolve to PROB is fine as long as PROB is genuine and not market manipulated. If a market is assumed to resolve to a definite conclusion in a timely manner, but contrary to that expectation, it doesn't, then there should be no problem. The PROB value is a good (albeit not as good as correct answer) amount to resolve to, rewarding contrarian opinions. OTOH if there is collusion with the market maker by another party (or colluding with himself) then this is a problem. So throw out the baby? No don't throw the baby! That problem is no worse than the incorrect resolution problem, and the incorrect resolution problem is not one MM cannot possible solve without substantially changing the ethos of the platform. If someone is going to rugpull with a PROB auto-resolution, they'd just do it with an incorrect resolution for more profit. Therefore: Auto resolve to PROB is fine. No need for extra complications for last-second-bet prevention. If you want to add a protective mechanism, then simply fine someone M$50 if their market gets autoresolved. Also, don't allow them to create a new market until all markets that need resolution are resolved. This would in addition require an immutable resolution date.
@NcyRocks, that sounds like a smart compromise. So the creator still chooses a single deadline for the market and trading stops exactly then, but they still have some standard window to resolve it before it auto-resolves? If so, I think that window should be either an hour, a day, or a week, for Schellingish reasons. > you can add liquidity directly now! Oh, nice! Just added M$100!
I think that a delay between the closing and resolution would introduce some ambiguity as to whether the market is actually going to resolve to PROB or be resolved by the creator, which would solve a lot of potential manipulation problems. (@dreev, you can add liquidity directly now! It's been implemented for a few days. You need to click on the three dots next to the volume (on desktop; don't know what the mobile layout is))
Oh, right, I stand corrected on the part about not expecting the market to be abandoned. But two more counterpoints: 1. Maybe the creator was in fact waiting till the last second and maybe your last-minute trade nudges them to actually resolve it. Especially if you seem to be doing so manipulatively! 2. Why wouldn't other traders correct the market price? Especially since participants in the market care about it resolving to the truth. Ultimately we could call this an empirical question: would auto-resolve-to-PROB result in unsatisfying shenanigans or not? My gut says not, but now we've got a more grounded prediction we can hash out in a new market, assuming we agree on auto-resolve-to-something...
Bearing in mind that this is the absolute final auto-resolving deadline, not a closing deadline prior to resolution later, I think 10 minutes before it, I will be pretty confident that the market is abandoned- it'd be running a pretty high risk of something going wrong for a market creator to delay resolution that close to autoresolution on purpose. Very different from ten minutes before a regular closing time. I could then move the market arbitrarily and gain from being more "right" about the probability than existing traders, no one would have incentive to correct it, and if anyone did I could simply move it back (and they'd lose money by "being wrong" and I'd make money from "being right", presuming it did ultimately resolve at my preferred value).
@jbeshir Ah, great that we mostly agree! As for resolve-to-PROB vs resolve-to-N/A, I think you're saying that even though the quiescence rule eliminates actual sniping, a sniper can still make a crazy trade right before the deadline and no one else will be incentivized to correct that. But isn't that only true if you actually expect the market to be abandoned? Before the deadline hits you don't expect that. I suppose you could make small trades before the deadline to keep the market alive long enough past the deadline that the abandonment probability gets high enough for your strategy to be lucrative. But that's starting to feel far-fetched to me, or at least a bridge to cross when come to. In any case, auto-resolve-to-N/A is most of the way there from my perspective! Definitely better than the status quo.
On the other hand, PROB with quiescence would I guess let me pick what the market resolved to, since I probably have deeper pockets than anyone else screwing with it and my limited patience to play last-minute games is no longer a blocker, so it -does- have that to recommend it! 😁
(Post-quiescence doesn't fix this because of the no-reason-to-move-a-non-resolving-market-to-true-probability problem. There's no reason it'd be profitable to move it to the true probability rather than moving anywhere else once you were now quite sure the market wasn't resolving.)
Support in principle but don't agree with "PROB", specificially- I think N/A resolution is much preferrable. I believe PROB resolution is vulnerable would incentivise lots of weird pushing the market around at the last minute. For example, I just bought 41 M$ if NO in this market for 4 M$, moving it to 0.7. If it resolved PROB I would now receive 12 M$. If I had pre-existing NO I could have increased its value by much more than this. In the general case, you can make bought shares worth dramatically more by pushing the current odds of the market maker around at the last minute with comparatively small amounts of money and with no actual resolution expected there's no reason for other people to expect to make money trading it back to accurate probability. For this reason I strongly favour N/A resolution if the market creator doesn't resolve after the final deadline. With N/A resolution unrolling the entire history of the market, nothing you did in a market resolved N/A matters, so rational betting behaviour is identical to assuming/conditioning on resolution happening.
I just bought a bunch of both YES and NO to subsidize the market. Also we were just talking about a nice case study in the Discord. @Adam nudged me to resolve one of my closed markets (about crime rates). I had been procrastinating on doing so because I'd hoped to do more research in order to determine a definitive resolution. But without a hard deadline, I was failing to ever actually do it! (And I did end up resolving-to-PROB since that was the best estimate for what I'd've determined if I had actually done the research about crime rates needed to get the definitive resolution.)