Will partial resolution of yes/no markets be supported by 2022-11-01?
29%
chance
This market resolves Yes if Manifold adds a feature to support partial resolution of yes/no markets by close. It resolves No if not.
Suppose there is a market "How many times will Donald Trump tweet in 2023?", which resolves to 0 if there are no tweets, 100 if there are 100+ tweets, 50 if there are 50 tweets, and so forth. When the market opens we know it will resolve between 0 and 100. If Trump tweets 50 times in August then we learn that it will resolve between 50 and 100.
Partial resolution would allow the market creator/resolver to set the minimum resolution to 50% and immediately pay out 50% of "No" shares. The market will continue to be open for trading in the 50-100% range. No doubt this will take some fancy math.
If there was a market for "what proportion of the Spice Girls will be alive in 2050?", then if one of the Spice Girls died, this could be partially resolved, setting the maximum to 80%. So we could have partial resolution in both directions. Though perhaps not both at once, depending on how hard the math is.
An example market that could be partially resolved to 50-100%: https://manifold.markets/MartinRandall/will-existing-markets-be-converted
Benefits: improves prediction accuracy, reduces fear of improper resolution.
Costs: user confusion, opportunity cost.
Unknowns: how to handle liquidity, how to handle N/A resolution.
Small print: I will not trade in this market. I will handle edge cases according to my best judgment, feel free to ask resolution questions in comments.
Austin bought M$10 of YES16 days ago
@Yev hm what's your leading proposal on scalar markets? I think we're currently leaning towards bucketized, exclusive multi-category CFMM markets to represent scalars
Yev bought M$100 of NO16 days ago
This feature is only useful if you use PROB as a poor person's scalar market. Actual scalar markets (with partial resolution) will be make this feature unnecessary.
Austin is betting YES at 50% 17 days ago
One worry - with partial timed resolutions, the specific resolution time starts mattering a lot. Eg if the resolution is weekly, and a partial resolution was supposed to happen on 5/20, but the market creator is on vacation for 2 days and the protest ends on 5/21, then people can get rewarded for jumping in on that.
Maybe you could resolve this using "who owned the shared as of exactly 5/21"? Marking each share with a timestamp might be kind of annoying, but our ledger should be able to reconstruct the state of ownership as of any point in time without such timestamping
Austin bought M$20 of YES17 days ago
Oh that's super cool! Reminds me also of my proposal for trendcasting https://manifoldmarkets.notion.site/Proposal-Trendcasting-36321d16185543ad93d6e9484f8718c4
Another use case: a "how many days until X (eg, the protest ends)" market, which resolves 0-100% based on the number of days it takes. That can have an increasing 1% partial resolution every day until the event happens.
Austin bought M$5 of YES17 days ago
Oh dang, this is really interesting! I think the math works out okay - since every YES + NO share was minted by paying in M$1, we could just send out the M$0.5 to each NO holder, and then now a NO share represents the chance it resolves to 50%. Need to think a bit more about whether the incentives are still aligned (if I had bought YES shares at 33%, what does it mean once the floor is raised?)
One more benefit that is actually quite important is that it reduces the amount of time that bettors have their capital locked up - faster resolution means that
One more idea I've wondered about, similar to this: Continuous resolution of markets e.g. you could have a prediction on "How many DAU will Manifold have by the end of the 2022", but actually pay out the market every week. This might work better for Free Response markets where you have buckets for scalars, because the continuous resolution solves the DPM problem of later traders capturing gains of earlier traders https://kevin.zielnicki.com/2022/02/17/manifold/