Thread

depends on what you're talking about. are you looking at how it's discovering relays to connect to or are you talking about the relay chart at the bottom of the dashboard?

Im talking about the relays you query for events to calculate users and other stats, these relays are coming from nip-65 on popular relays right?

It’s measuring something that is not human users. That’s why the retention is “low”. There’s lots of reasons for ephemeral keys for example.

Check out the tooltips. Lots of the charts filter out ephemeral pubkeys that have only been seen on events like gift wraps or marmot messages. Basically any time it’s a throwaway key we don’t include for aggregation stats.

it's ok, inflating the numbers is good for marketing purposes

10/4 I used the word “ephemeral” a little loosely there. Temporary probably made more sense there. NIP-17 on the brain.

Yeah, we're talking about the same thing. NIP-17 uses kind 1059 - we don't pubkeys that publish 1059 events since we'll only ever see those used one time.

Sorry, what I meant was there are many cases where throwaway keys are used which are not scoped within NIP-17/Marmot/etc and would still fall through into your w0 cohort I also suspect there’s lots of “probing” going on- people making bots and testing if they get past certain relay/WoT filters for example. I think a good chunk of this can be filtered out by simply ignoring any key that doesn’t last more than some minimum amount of time. I proposed a month. No idea what the “right” number is. But this would get a more accurate reflection of retention than the current methodology.

Isn’t that just gaming the retention figures? I agree with the idea that we can probably come up with something that cleans this up a little, but a fair way isn’t immediately obvious to me

Not obvious to me either tbh. It’s not clean on an open protocol. I don’t think it’s gaming if you state the methodology clearly and even the reasoning behind it. Maybe display both versions and let people interpret themselves. Right now you could say it’s already gamed in the other direction. We don’t need to footgun ourselves IMO - it’s completely new territory with limited comparables.

W0 baseline is incorrect not measuring human users, likely measuring something else as well in that cohort

Like Smart fridges 😂 View quoted note →

Lol exactly

New users is irreelevant metric where you can generate a million of them between breathes. I guess DPU w/ profile+followers is the only somewhat relevant metric even tho that can be easily manipulated

exactly, unless you use a WoT service provider like Vertex to determine the reputation of the user, which is determined by other people that already have reputation.

That's fair, but it's also fair that human users aren't the only users that we actually care about on nostr. DVMs are users. Servers are users. And yes, even bots are users. This is more of a question of what do we want to measure vs are these numbers correct. Would love to talk through this more sometime though because I'm sure there is some fun stuff we could do to build a chart that would allow us to visualize these breakdowns better.

That's a good take. Yeah the best as you said would be to categorize these users, giving the option to remove from the stats the non-reputsble according to some threshold and ranking algorithm. Let's continue this convo on Signal

Agree. The drop off rates make no sense. Wonder if ephemeral users/keys are being counted somehow.

For retention stats we gotta differentiate humans. You’re right we care about other activity too, but showing retention flatlining after a week isn’t fair to the ecosystem. Nobody can show that graph to someone on the fence and expect them to try it out, let alone show it to a potential investor in a product developed here.

Interesting... some user stats here definitely seem unrealistic

Good call! I don't have anything like that yet but I'll add an issue on the repo to try and get it added sometime in the future.

Thank you so much for the Zaps, Jeff! Im really glad you liked my comment. Your initiative to bring more transparency to Nostr is truly inspiring, and I’m happy to be a part of this conversation. Keep up the amazing work!

not bad

all them washtrades on the Nostr are too damn high!

This is the most reliable skin in the game stat on the page

😂

Nostr users? Hahaah

Until you realize it’s just the same 270 sats zapped around a few hundred thousand times 😂😂

🤣🤣

Imho I don’t think it should be included at all… They aren’t in nostr lol

I have it running on a bare metal server that costs ~$70/mo from Hetzner.

ahahahaha

There is - it's about 5k urls, many (most) of which are not live. I can have a look now that it's been running a week or so to see what data I have on those. The relay stat tracking code is very basic at the moment, it's an area of the app I'd love to improve on.

Got it, so approx 5k relays tracked inception then? I’m curious what the event/relay distribution looks like. Not so much from a decentralization perspective as much as a behavior perspective. Assuming Damus and Primal rule, do other relays play a role and to what type of user.