https://manifold.markets/stats
An engaged user is a user who has traded in, commented on, or created a question on at least 2 out of 7 days in each of the past 3 weeks.
๐ Top traders
| # | Name | Total profit |
|---|---|---|
| 1 | แน5,493 | |
| 2 | แน3,975 | |
| 3 | แน3,617 | |
| 4 | แน623 | |
| 5 | แน235 |
I'm curious who of chis, pc, vlovic, and marcus were:
1) doing some models solely based on the numbers on /stats
2) explicitly calculating the number of 'engaged users' based on api data for individual traders
@jacksonpolack I'm silly for not realizing that was even possible. It makes all the YES bets so much more reasonable.
@SophusCorry Yeah. That was what Chris was implying with his comments the whole time as well, also why he was so close to the actual numbers both times. If you look at his bet history, he started by taking the gamble himself, and then shortly after went for the YES'. It's pretty funny to read his comments now. Well played. Not that their play mattered too much for us, given that our gambles were made without the thought of using API access anyways, so we would have made the bets either way.
But in hindsight, it's so obvious that the API could be helpful here. And the mis-matched odds should've made us more suspicious.
@JamesColiar we live and we learn. At least it's just mana. If we didn't improve our prediction heuristics and strategies the whole thing would be a waste of time. But still fuck.
@JamesColiar Getting the exact number is impossible I think due to activity on private markets not being accessible via public APIs, hence why my numbers weren't perfect.
Because the engagement metric is almost entirely about the past, as of about 36 hours ago it became inevitable that engagement would exceed 1200 even if everyone stopped using Manifold entirely. So betting here was just about trying to get the best price betting on a mathematical certainty. Sorry ๐ฌ.
This spike up to 1200 was inevitable, but there is less certainty ahead: peak engagement will be somewhere from 1242 today to 1652 on Aug 22nd, depending on exactly which users are active in in the coming days.
It is hard to gauge from retention stats exactly how this will go. On the one hand, DAU numbers of accounts that were not active before LK-99 are impressively resilient:
And there is a dedicated cohort of new users using the site literally every day with high retention:
But the intermediate level of dedication, users that are active at least 2 days out of every 7 - which is the one relevant to the definition of engagement - has fallen by about half in a bit over a week:
So I suppose we will see.
@jacksonpolack I wish I was using the API. Just models that have lots of uncertainty. I would guess it is only Chris on the API atm
@chrisjbillington Is the fact that it became inevitable because of the way the weeks roll over, or was it some other reason?
@DanielTilkin Yes. Active users were rolling from the second most recent week into the third most recent week in the three-week "engagement" window, increasing the number of users that were active in two days in all three weeks by a sufficient amount that engagement was on track to hit 1200 without any further active users in the most recent week in the window.
@chrisjbillington So, Jul 31 / Aug 1 had something like 3000 signups, 2000 activations. If 20% become engaged, then you get 400 new engaged users above the 800 exiting. And they start entering into their 3rd week on Aug 14 / 15
This question is interesting to me because I'm trying to understand how many of the superconductor users become engaged users
@JamesColiar I did suggest the possibility of them doing this. I just didn't imagine the retention from the august 1st users was going to be as good as it was. We have 3.4k weekly users and a lot of them stopped using it in the last 7 days likely, so only like ~2.5k people still using the site.
@MarcusAbramovitch In about an hour. They're usually delayed by about two hours in relation to loan payouts.