🏅 Top traders

My current credence here is 60%.

Again, I want to reiterate that I don't think this is the best solution - I think it would work better if Manifold moved away from authors having total control over resolutions, as has been discussed in the comments. But thinking about just this change compared to the current state of Manifold, I think it is a bit more likely to be an improvement on net. See my previous comment for the reasoning: https://manifold.markets/jack/would-it-be-a-net-benefit-if-for-mo#n12CgKQ48UafkExG9Qjy

@jack I feel like an important consideration here is that there are many less-restrictive options for improving market reliability other than broadly disallowing this. Examples:

Flagging markets that the creator has traded in

Karma systems

Appeal systems

Enabling users to bindingly disavow trading in their own market

Enabling users to bindingly nominate a third-party judge

Given Manifold’s impressive calibration notwithstanding not having this policy (https://twitter.com/ManifoldMarkets/status/1589703623935565826) it seems like most of the benefit would be to aggrieved shareholders, not public epistemics. And public epistemics would be harmed from the dramatic reduction in the number of markets.

@Radicalia Your fourth suggestion, “Enabling users to bindingly disavow trading in their own market” is what this is about, except possibly opt-in versus opt-out. I think the main difference between those would be what users do on their very first market, and I think it’s better to encourage users to do the safe thing on their first market, especially since the one of the biggest concern for rugpulls is users that haven’t created markets before.

@Gabrielle But that’s visible (1st market) from user profile info on the market page, so the median market predictor should have what they need to account for that, no?

@MattCWilson Currently predictors often avoid new users because of the threat of a rug pull. If this setting were enabled, predictors could feel safer betting in the markets of new users, which helps encourage new users to create more markets and continue using Manifold.

@Gabrielle I'm not sure markets of new users are being avoided. The first market I resolved had 115 traders (but maybe I would have had even more?).

@Gabrielle I would be interested in testing that hypothesis; a user-community survey of established users about their practices in betting in new-user’s-1st-markets would answer it.

If I create such a survey, would you be open to subsidizing a market on the results for, say, 1000 Ṁ ?

@MattCWilson Sure, that sounds reasonable. I’ll do M$1000.

@Gabrielle Awesome! How about I draft a Google survey and invite you to give feedback? I think I see you on the Discord; I’ll DM you there.

@MattCWilson Sounds good, I’ll follow up with you there.

@Radicalia Definitely, I also mentioned this in the market:

Note: I would rather move away from the model of author having full control over resolutions, towards one where authors resolve most questions but in cases of dispute the final authority rests with a resolution council of some sort. Will it be possible to dispute and override market resolutions by end of 2023?91%‌ . But if we're still stuck with authors deciding all resolutions, it is possible it may be overall good to not allow authors to trade.

For the purposes of this question, I am assessing the change to author trading assuming Manifold doesn't make other relevant feature changes.

@Gabrielle If it was mandatory for the first market, which I believe would plausibLy be a good idea, then it still would not capture “most” markets per the title.

If you can opt out, then I’m not sure that it’s accurate to say the creator is “not allowed” to trade.

@Radicalia In the description for this market, Jack wrote

What I'm thinking of here would be something like if markets default to not allowing author trading but it can be optionally turned on.

So this would be an optional feature that a market creator could choose to use or not. I don’t think anyone is advocating for it to be mandatory in any case.

I see this as a helpful tool for market creators who want the ability to safely precommit to not have the temptation to bet in their own markets.

@Gabrielle Thanks for pointing that out. I would just note that if that is sufficient for @jack to resolve yes, then IMO the title of the post is pretty misleading insofar as that type of thing is not really about anyone being “allowed” to do anything and would not on its own be about “most markets.”

@Radicalia What Gabrielle said is indeed one of the possibilities I was suggesting. The idea is that each market clearly indicates whether author trading is allowed or not, and if most markets do not allow author trading then traders will tend to have a higher degree of trust in those, and can avoid markets that they trust less.

I think this is indeed a proposal that fits with the title. It is about what people are allowed to do - even if authors can choose which ruleset to opt in to, once set the rules are binding and communicated clearly, and that's important. And the hypothetical system under consideration would be assumed to apply to most markets for this question.

Ok, here's an update with my current thinking.

I still think it has a lot of very important upsides for increasing trust in resolution accuracy.

And I still think it has a very important downside of discouraging some authors from creating markets. I think many authors will still make markets, but it will decrease authorship significantly.

A recent development that is quite relevant is that Manifold has finally implemented the ability to undo wrong resolutions! https://manifold.markets/jack/will-it-be-possible-to-undochange-t#ADaZGoVhPXHhH3CQg98s. This means that at least for objective questions we no longer have to worry much about misresolutions and blatant fraud. Allowing the author to trade on a very objective question (e.g. Will Biden win reelection for President?) seems to be very low downside compared to the upside.

For subjective questions, I think removing author trading to reduce author bias and increase market trust in the author's decision is still very important, and I think having an option that authors are encouraged to use for subjective questions would be helpful. Not clear whether most markets are subjective or objective though or which way the default should go though. Glancing through the markets, it seems close-ish to a 50/50 split between objective vs subjective questions.

@jack I don’t think looking at author discouragement separately from increasing trust in Manifold. If I think about where Manifold is ten years from now, if it still exists then it will need to expand its user base by a lot (or completely pivot). I think that expanding is really limited by user trust right now.

There are a lot of users who sign up for an account and then stop using Manifold in less than a week (somewhere around 70% of new users), and my theory is that a lot of these are because they interacted markets that seemed “sketchy” and got discouraged and left. Maybe even a few left because they created markets and not many people invested (because people didn’t trust that the markets made by new users would be resolved correctly). And of course some of it is users that just get bored or decide they don’t care about Manifold, but I think a decent portion are discouraged by a vibe of “sketchy markets”.

If Manifold becomes much bigger, any decrease in authorship can be far outweighed by the number of new users, and I believe that the trade off with an optional switch is far more valuable to helping user retention than it will cost in potential market creation.

@jack My concern about the subjective/objective split is that even a lot of markets that seem to be objective wind up having criteria that are subjective when applied. A relevant example is the Texas blackout market, where users would probably have thought it was objective (and safe for the market owner to trade in), but then it wound up being really subjective. Some markets are absolutely subjective (“Will I be convinced of…”), and some markets are absolutely objective (“Will this stock on the NYSE reach this value before this date during daytime trading”), but it seems to me that the large majority wind up in a weird middle state that aren’t quite objective even when they’re intended to be. That middle state is the most dangerous for market creators, since they’ll think it’s safe to trade in their own market and then find out that it actually was a subjective question and be biased when making the final call.

For one example, my market on the Department of Justice’s case again Google (https://manifold.markets/Gabrielle/will-the-justice-department-win-its) was intended to be objective, but it plausibly will be a weird middle case where I’ll have to define whether a settlement is more of a DoJ win or a Google win.

@Gabrielle I really don't think there are a lot of sketchy markets right now. There are only 5 (and there a lot of markets total) on this list and one should not even be on there: https://manifold.markets/enemel/what-markets-were-intentionally-and-56c95e8bf5f4

Furthermore, most markets you bet in will not resolve in the first week, making your theory unlikely. I think people leave in the first week, because they realize most markets are on longer horizones (e. g. elections).

@kottsiek I see a lot of comments (especially on Destiny markets, but also otherwise) where new users claim that a market is going to resolve incorrectly because the market creator has shares in a particular direction. I think that seeing these markets scares users away even if no market ever resolves incorrectly. They give Manifold an aura of being too incestuous, which users want to avoid, especially if they might otherwise want to spend money on the site.

@Gabrielle

I think that expanding is really limited by user trust right now.

I generally agree that this is important which is why I am very happy to see the change to allow fixing wrong resolutions. Now that is out of the way, I think the remaining issues are much less bad. Subjective markets are already prone to resolution disputes anyway.

There are a lot of users who sign up for an account and then stop using Manifold in less than a week (somewhere around 70% of new users), and my theory is that a lot of these are because they interacted markets that seemed “sketchy” and got discouraged and left.

I don't think so, I would think that's a pretty typical rate of people not sticking around for any new site. For one point of comparison, I just googled for twitter user retention rate stats and apparently "over 60 percent of new Twitter users fail to return the following month". I suspect interactions with sketchy markets are responsible for a small fraction of users leaving.

My concern about the subjective/objective split is that even a lot of markets that seem to be objective wind up having criteria that are subjective when applied.

Yes, I brought up this point too on one of the other markets. I think removing the author trading helps significantly but ambiguity is still just inherently a problem, and that needs to be addressed directly with one of the many other approaches that have been suggested (delegating resolution to experienced and reliable resolvers, helping users draft good questions, etc)

I would go so far as to say that market creators should be unable to see who bets on their markets. (This would be easy to evade, but it would still be valuable as a symbolic line, and it would give market creators plausible deniability if they wanted to resolve in a direction that might be socially costly for them.)

One other point I want to make is that even if there were no net bias in how authors resolved questions that they had made bets on, the bets would still move the resolution toward being disconnected from reality, because it would mean authors would be overcompensating for their own bias and misresolving questions against their own interests equally often as they were undercompensating for their own bias and misresolving questions in favor of their own interests.

Prediction markets are a good world-modeling tool when they're "transparent" in the sense that thinking about how to profit mostly involves thinking about what will happen in the parts of the world that they're ostensibly about, instead of thinking about the market creator's psychology and incentives or quirks in the resolution criteria or the mechanics of the market. Being able to look at a seemingly mispriced market about ducks and have my first thought be "huh, I wonder whether I'm wrong about ducks" and not "huh, I wonder how I could get screwed over in some unexpected way despite being right about ducks" would feel like being able to breathe clean air for the first time. I think Manifold selects for people who enjoy or at least don't mind all the extraneous trickster nonsense, and ends up underestimating how off-putting all this is to people who are here out of an interest in forecasting the actual world out there. "Don't have the same person be the bettor and the judge" just seems like an obvious common sense integrity boundary.