Suppose there are markets on whether or not a policy will be beneficial according to some measure and the policy is only enacted if the market probability is above 50% and otherwise resolves N/A.
There might be issues with the market referencing itself, allowing self-referential "manipulation" and issues with not enough liquidity for probabilities to be priced well if the market is in certain states (EG market N/A probability is very high)
This is a test to see if issues will occur in a toy setting.
Rules for Market Resolution
I have selected a hidden random time within April 1-5 to stop market trading and resolve the markets.
There are a set of 5 submarkets: P=0.2, P=0.45, P=0.5, P=0.55, and P=0.8.
For each submarket, there are three options:
- YES "policy will be enacted and beneficial"
- NO "policy will be enacted and not beneficial"
- N/A "policy will not be enacted"
Say for each submarket,
Y = market price of YES at close
N = market price of NO at close
If Y/(Y+N) > 0.5 at time of market resolution, resolution will be:
YES=P, NO=(1-P), N/A=0.0
If Y/(Y+N) < 0.5 at time of market resolution, resolution will be:
YES=0.0, NO=0.0, N/A=1.0
Related questions
🏅 Top traders
# | Name | Total profit |
---|---|---|
1 | Ṁ67 | |
2 | Ṁ66 | |
3 | Ṁ39 | |
4 | Ṁ24 | |
5 | Ṁ21 |
In order to resolve the P=0.5 market, I used the API to get fine-grained probabilities.
P=0.5 YES looked like
{
'pool': {'YES': 173.00000000000006, 'NO': 93.73002681220085},
'probability': 0.3514041067382122
}
and P=0.5 NO looked like
{
'pool': {'YES': 173.00000000000009, 'NO': 93.73002681220085},
'probability': 0.35140410673821215
}
It looks like the probability of P=0.5 YES is a litttle bit higher coming down probably to floating point errors.
@JamesBakerc884 I'm not 100% sure, but first note that these are unlinked binary markets, their resolutions don't need to depend on each other.
Secondly, I think but am not 100% sure that the way it works is that if a single binary market gets resolved to 0.3, holders of YES in that single market get 0.3 and holders of NO in that single market get 0.7. This might sound a little confusing because I have YES/NO labels on many of the single markets on this page, but note there are two different uses of YES/NO here.
Hm I think a more robust resolution mechanism would be Y/N being determined by an average, median, clipped average, ? over a time period instead of a random time
I’m surprised people are buying P=0.8 YES upwards, as it should be easy to buy it down to 0% just before market close and have it resolved to 0. (Submarkets don’t resolve to N/A, they resolve to 0.)
@ms_test The markets are resolved at a random time in a 5 day interval so it might be tricky to time this. I encourage people to try doing this if they want.
@ms_test My answer is, "because this is a toy, test market" and I'm willing to see if what "should" happen actually happens.
And I don't know Manifold well enough to know what the actual incentives are. This 3-submarket thing, where you can bet 1 or 0 on three things independently, seems quite complicated.
If it's easy to game P=0.8 to N/A, then wouldn't it be even easier to game every other submarket?
@NoaNabeshima not anyhow more clear. Yes holders? Yes potential payout? Yes - amount of answers which have >50%? Yes - percentage of the YES option?
@KongoLandwalker Yeah sorry I think this wasn't clear:
For each submarket,
Y = market price of YES at close
N = market price of NO at close
I updated the description. Does it seem good/clear yet? I like this feedback.