Manifold in the wild: A Tweet by Michael Wheatley (0/100 Completed Projects)

@cosmojg @goblinodds As a manifolder, you should know better than to think this is settled science. https://manifold.markets/MartinRandall/is-the-sky-blue-resolves-to-mkt?referrer=Botlab

Oops, a bot just showed up and moved the probability but it was at 90% for over 24 hours!

you checked on the API?

@MichaelWheatley I'm not sure exactly when @MatejMecka bet, but around when we were first talking about it potentially being 24 hours, it looked like there were still 40m left.

@MichaelWheatley You can see timestamps by hovering over the "X minutes ago" blurbs:

But it doesn't actually matter when MatejMecka bet since they didn't change the market price by a whole percent (note Martin's closing criteria in the description about that).

Which I believe means this should resolve to 90%. That's what I was assuming would happen when I decided to stop changing the price.

@dreev yup, I did the math on the API timestamps just now and its over 24 hours.

Yes, I think there has now been a 24 hr gap. I suppose I should confirm with the API.

@MartinRandall A bird tells me that the UI says "a day" but is rounding up. Bettor beware.

I guess I could make like M$1000 by fighting more over this price but also I guess I'm ok with saying 90% is a pretty fair answer to how true it is that the sky is blue, so maybe I'll call this good, let it resolve, and then I can deploy this capital in the bitterly contested question of whether Biden was the president last week!

I currently have zero shares so I can be scrupulously fair during the transition to the Poisson process. I am targeting a 100 day average resolution time if the market dies not settle before then.

@MartinRandall Thanks for doing all the work on this! With 100 days in expectation, I think over a year is still pretty possible? Which would feel kinda painfully long maybe?

@dreev Is there a parameter setting that makes less than a day and more than a month both quite unlikely? Let me know if a graph of some kind would be helpful; I'd be happy to nerd out on this.

I wrote up what I'm thinking for my Biden market here:

https://manifold.markets/jack/will-biden-be-president-on-915-reso#Kw05SzzVLw7jKREuugpM

@dreev I think I'm going to do something much longer than Jack's market just to gather more data on how these markets (mis)perform.

If bettors can predict where the market is likely to settle in 100 days they can let it settle there now and have the market resolve.

So how is this poisson process going to work?

@MichaelWheatley I think I'll need a source of public randomness and then I can do the same trick where anyone who notices that the random generator has spit out a 99.9% probability can comment and the market will resolve at the probability as of that time.

@MartinRandall The method I'm using on some of my markets like https://manifold.markets/jack/what-is-the-probability-that-a-coin is to use https://avkg.com/en/daily-random/ as a daily public randomness beacon, to close on a random day in a range. I'm not going to go below day granularity if it's a manual process, if it were automated then it would be pretty simple to make this better.

If you want other more fine-grained sources of public randomness, check out some ideas on https://manifold.markets/jack/resolves-yes-1-chance-na-99-chance

@jack Do you know when new avkg numbers are published?

@MartinRandall I don't, that is a potential issue I guess.

This is a public randomness beacon I'm considering using if I can put something together on the API to make it automatically close: https://drand.love/.

I am trying to avoid the situation where the coin flip says the market has closed, but people don't know about it and keep betting.

In order to avoid this cluttering up the "resolving soon" list, and to reduce the chance of my actions in this market causing it to be biased in either direction, I am thinking of extending the close date, but stipulating that anyone who notices a 24 hour gap in trading can post a comment to that effect, and I will then resolve the market at the probability at the time of that comment, after checking that the gap exist. If I notice such a gap I can do the same. Does anyone have any comments on that? I especially welcome comments from people with a high number of shares in the market.

@MartinRandall That sounds reasonable.

@MartinRandall Yeah, agree this seems fine. I think most likely there will continue to be at least 1 trade per 24 hours on both our markets until we start doing the random process to determine the close time thing.

@MartinRandall So if I see that no one has traded in 24 hours, I can buy 10000 NO and comment, and you'll resolve to the probability after my bet?

@Yev No, before your bet. Though that would be funny.

I need to decide what to do if there's a 24 hour gap followed by a bet instead of a comment. Probably resolve it.

@MartinRandall You wrote "resolve the market at the probability at the time of that comment". I guess it should be "resolve the market at the probability at the time of that gap"?

@Yev Yes, let's use that wording. Thank you for asking about the ambiguity instead of exploiting it, have a tip.

@Yev Oh, and I should resolve it if there's a 24hr period where the probability doesn't change (to nearest percent), so stabilizing the market with limit bets is an option and so we can't be trolled as cheaply.

@MartinRandall I didn't try to exploit it because I don't expect this market to fully stabilize, not because I'm nice. :)

@MichaelWheatley @jack @Yev @dreev changed resolution criteria as discussed, please let me know of any concerns.