Quirky market.
On the one hand, you have the Manifold market predictions on the linked ACX questions, as of the time I made my wagers (so, the numbers recorded in the post that comes out 1/9/24).
On the other hand, you have the BLIND Zvi predictions from that post. These were made with minimal reference to prediction markets, and did their best to exclude known market prices there, but did include looking at financial markets.
To avoid this being about places where I knew nothing or made a large conceptual error on something I know nothing at all about, since that's no fun, we will start by adjusting Zvi's numbers to within 20% of the market price. This impacts question 20 and 31 a lot, and question 21, 26 and 28 a little.
Also I flat out decided not to answer question 10 about the Mexican election, because I didn't find it worth doing the research, so that one is excluded.
AFTER that adjustment, this question resolves TO THE PERCENTAGE of weight that you should give to Zvi's blind numbers, versus Manfiold's market numbers, that gets the best possible Brier score, with a minimum of 0% and maximum of 100%.
All questions get equal weight.
I am invoking that Austin Chen shall arrange for a mod to calculate and resolve the market when the time comes. I trust us to find someone who can do the math.
Putting this up early on purpose cause it's more fun to have a blind period first, but you do know how often there was otherwise a >20% disagreement.
1,000🏅 Top traders
| # | Name | Total profit |
|---|---|---|
| 1 | Ṁ790 | |
| 2 | Ṁ14 | |
| 3 | Ṁ12 | |
| 4 | Ṁ6 | |
| 5 | Ṁ5 |
People are also trading
Lest people take too dim a view of Zvi's prediction capabilities from this, I simulated several thousand resolution possibilities with independent probabilities for each competition market given by splitting the difference between the adjusted Zvi predictions and the manifold predictions. Resolution percentage for this market turns out to be 0% or 100% about half the time. I invite the high-dimensional geometers to try to explain this for the rest of us.
I am no longer a mod, but nevertheless here are my calculations. I interpret ambiguities and unresolved markets in Zvi's favor, but I still find that this market should resolve to zero.
There's a possibility that there's still some kind of bug in the spreadsheet, but at this point I don't think it's very likely. Perhaps if it really has to be a mod to do the math, some mod can double check my spreadsheet and then resolve this.
How will you deal with the the US Election winner multichoice? The manifold market for this is a bit different from the Metaculus, it only has options for democratic, republican, and "other". Should all three of these options be counted as individual components that are weighted evenly in the Brier score? Should only "democratic" be, since that is the number you bolded in your post?
@BoltonBailey I believe I gave my probabilities I gave to Metaculus, if I didn't I'd be fine grouping all the Others together, this market isn't popular enough that I'm going to worry about this much, it's not going to happen.