It seems mad that in the diamond League right now, a score that would put you No. 1 in one cohort would leave you barely even in the top half in another. Obviously it's early days for this season, but similar effects have been observed previously, and remarked upon.
Will Manifold change the way users are sorted into cohorts such that cohorts in the same league are similarly competitive, by the start of February?
I'm going to make this a little bit subjective because I want to reserve the right to resolve in the spirit of the market, but here are some rules of thumb:
First place in one cohort should not be 8 times the size of first place in another cohort
Promotion in one cohort should not be 4 times promotion in another cohort
First place in one cohort should land you in promotion zone or "very close" to it in any other cohort
Demotion in one cohort should not be higher than promotion in another cohort
"Very close" (bullet 3) means within a few hundred mana, or 10%, depending on which seems most reasonable to me.
I'll mostly be looking at diamond, but I expect platinum to meet the above standards too. I think other leagues are a bit more unpredictable.
I'll allow any of the above rules of thumb to be broken if I think that an effort has been made and it has been mostly successful. There's always the chance one or two users get very lucky and I don't want that to affect resolution.
But also, if all of the above criteria happen to be met and it seems like coincidence and I have reason to believe no action was taken, I may decide that doesn't count.
I may resolve a few days after the start of February, just to make sure I understand the rules by which that season works, unless I get clear confirmation that it hasn't been addressed.
Market can resolve early if this is implemented. If leagues cease to exist or their structure gets changed such that this question doesn't make sense, I'll resolve N/A.
Since it's subjective, I won't bet in this market.
1,000🏅 Top traders
| # | Trader | Total profit |
|---|---|---|
| 1 | Ṁ283 | |
| 2 | Ṁ62 | |
| 3 | Ṁ60 | |
| 4 | Ṁ39 | |
| 5 | Ṁ17 |
People are also trading
Ok, I think this is a fairly clear NO. Looking at Diamond for S9, conditions 1, 2, and 4 were met, but condition 3 was not. First place in Slick Krakens was 8,497 and promotion in Bayesian Oracles was 12,161. This is more than "a few hundred mana" and also more than 10%.
I haven't checked Platinum. If somebody wants to make the case that the market should resolve YES (perhaps you have checked other seasons and you think the problem has been fixed but this season was a weird fluke that should be discounted), you have 24 hours. Otherwise I resolve NO tomorrow.
@Fion Interesting to note that 1, 2, and 4 seem to be easier to satisfy than 3. (See Canlum's comment below analysing a different season.) This might be a sign that I chose my criteria poorly (maybe I made 1, 2, and 4 too easy), but I'm not worried about it. I think if you can be coming first in one cohort but not even close to promotion in another (the leader in Slick Krakens would have come 6th in Bayesian Oracles) then that's a problem, and I think Manifold should work on improving it. Otherwise leagues are just a lottery of who you end up with.
In other words, the letter of the market is NO, and so is the spirit. Satisfying 3 out of my 4 conditions is not evidence that the problem is nearly solved.
Looking at Diamond for last month, it looks to me like conditions 1, 2 and 4 were met, and 3 was not far off:
1- highest first place was Shump in Optimistic Clairvoyants at 46681, lowest first place was Lion in Predictive Basilisks on 7302 (8x 7302 = 58416)
2- highest promotion was Nico in Serendipitous Medusas at 11370, lowest promotion was verymuchvix in Turbo Sprites at 3782 (4x 3782 = 15128)
3- Lion's 7302 first place is not within a few hundred mana nor 10% of Nico's 11370 highest promotion
4- The highest demotion was Irigi in Cunning Thunderbirds on 2164, so less than verymuchvix's lowest promotion.
@CanlumNodel Platinum is not close, however - I've not checked as thoroughly, but I don't think any of them were met.
@IsaacKing Whether it's a problem depends (among other things) on whether leagues matter. I think there's a strong case to be made that they don't, but let's put that to one side.
But I'm not convinced that you're right that people will self-sort into more balanced leagues over time. I think the grouping-by-interests prevents this. If there is a correlation between average profits and people's interests, then the promotions and demotions will always be biased by this.
There's a lot I don't know about how cohorts are determined. Maybe it's been a fluke that it's been so uneven so far, but it is really uneven.
Edit. Disregard this. I have now enacted changes. Detailed in reply (or check edit history of the description).
[I hope this is ok because so far people have only bet NO, but I'm going to change the resolution criteria slightly. I realised that actually I think they're already pretty much satisfied by season 4 of diamond.
Hold off on bets for a minute. Tweaking criteria. NO bets should be safe because I'm making YES resolution harder.]
I made the second bullet more strict, so mana required for promotion in different cohorts should be more similar.
I made the third bullet more strict, so coming first in any league should land you in promotion (rather than top 8). I'm allowing some wiggle room on this one, because you could have one cohort where users bunch up at the promotion cutoff and I think that's fine.
The new, stricter, criteria would be failed on all counts by diamond S4, although in some cases only just.
I doubt it, they haven't changed the payout structure of leagues yet either. on the one hand, cohorts by shared interests makes then much more engaging and worthwhile. on the other, you're right that it's a bit ridiculous how much certain overlapping interests are the dominant forces and mana making opportunities around here. I suggest that inter-cohort competition and rewards would alleviate some of this. if you're in the best gold group and get something for that I think that's much more interesting than diluted distribution