π Reliable nostr statistics now available at npub.world
Login to reply
Replies (61)
cc @The Daniel π , you seemed interested last time
Would it be possible for you to crawl any of the archival relays like relay.nostr.band or relay.primal.net to be able to have stats that go back more than two weeks?
I noticed them this morning, nice job π
That would not work for all statistics like total users, but more importantly, have these relays saved all the old replaceable events?
Ah you got me :)
Between pushing the feature and recording there was 1h. Anyway, thanks a lot
One suggestion. Rather than the "kind", label things with a plain English explanation of what they are. It would make it more accessible to non dev users.
suggestion taken!
I didn't want to add them because naming is kinda ambiguous, and it would have made the labels name longer, which is a problem on mobile.
Tooltips!
yes, but a tooltip on the legend is not possible on the charting library I am using. But a tooltip on the point and line is doable!
Given the massive disparity in follow count between all sources, I'm wondering how they are calculated here?
This removes bots and users who where never active I believe.
Other sources like nostr.band and primal seem highly inflated.
Massively inflated. Just curious about the pruning mechanism
It's using this. @Vertex
you can create a nostr note for the json :monoue:
suggestion taken!
Basically, I don't consider users that haven't be added to my database.
My program runs 24/7/365 listening to events. The moment you are followed by someone that is highly reputable and, you get added to the database.
So, from the 43M pubkeys reported in nostr.band, I only consider the 380k (supposedly real) ones.
The social graph is based on the Pagerank algorithm and the goodnpubs left are those who pass a minimum threshold rank
t *Y* 4 thread/
Maybe just a legend on the page for 85iq folks π
Great job, love to see it! Any plans to expand and include reliable stats for every NOSTR event? Primal cache seems to undercount reactions from legit accounts lately.
Yeah, I looked, after you mentioned it in my thread, and saw them.
Reliable stats would be very nice.
mmm, possibly
Cool, looking forward to giving it a full demo. One question so farβ¦how are you obtaining follow counts, and what accounts for the major discrepancy between npub.world and nostr.band?
ok, to calculate nutzap you have to search for 10019 kind and then search for 9321
This is the reason
View quoted note β
Onwardβ‘
Thanks, that makes sense. I always just assumed half of my followers were long gone. I built an app to remove the dead npubs that Iβm following, but thereβs nothing I can do about the count my clients display.
Plebs vs Zombies is a great tool, that also improved my ranking algorithm. If people curate more their follow list, then the quality of the signal increases
Thatβs an awesome endorsement, thanks!
that's another good idea, thanks!
Love it! Consider making it less "in your face" on the landing page. I enjoyed the clean search interface and use it a lot.
Uptober apply for Nostr too π
View quoted note β
Bravo!
thanks Gigi. mmm I am not sure, I like these little cute cards
Thank you Svetski
One other recommendation. This graph is less useful, just due to the magnitude of difference between total and active users. Could they be broken out into seperate graphs?
click total and it disappears
Thank you π€
Will you be able to track total zaps and total events etc?
total events by kind are tracked already.
If you mean zap amount, that is quite challenging as it requires a notion of lightning server reputation.
Ok. Hope you will achieve it too. :-)
What is your definition of "highly reputable"
Someone with a nice face and a fat sack?
π€£π€£π€£
higher than average pagerank
What is page rank?
Sorry Iβm asking a bunch of questions.
can you add an indicator to user's npub. world profile to indicate whether they are considered a real user?
this account is tiny but im 100% a real user and i'd like to know whether i'm recognized as such
There is no clear cut answer to whether someone is "real", of "fake". The whole concept of npub.world is that the user decides it, with the information presented in the page.
Take this example.
There are two odells, with the one in the middle being the clear impostor, that btw fooled many other (cc @ChipTuner @HolySats @roysterβ‘οΈ ) probably because they use clients that make the two profiles indistinguishible.
If I had a metric that says if someone is real or fake based on followers, if you would be real, then the fake odell would be real as well, and vice versa.
The information presented should be used in relative terms, meaning between the two odells, one is clearly fake.
There are two odells, with the one in the middle being the clear impostor, that btw fooled many other (cc @ChipTuner @HolySats @roysterβ‘οΈ ) probably because they use clients that make the two profiles indistinguishible.
If I had a metric that says if someone is real or fake based on followers, if you would be real, then the fake odell would be real as well, and vice versa.
The information presented should be used in relative terms, meaning between the two odells, one is clearly fake.
Understandable. its a hard problem, but there must be a better solution than having someone "reputable" follow you.
i havent put the time in to understanding pagerank yet but it smells of centralization, if you must be followed by someone of a sufficient reputation to be considered reputable yourself
Yeah this happens a lot. You're exactly right, primitive clients (like the one I use) don't even show follower count or "trusted connections" 2 things would be, I either added it to my list a few years ago, and/or followed based on someone else repost.
The other issue is this usually when resolving @ mentions, actually selecting the right npub. I don't remember many public keys, just the ones I have to juggle regularly. "
off to see how many odells I follow XD
Good work! But, difficult on mobile... Have to tap on specific data points to get x/y values, but it's super sensitive to position
Agreed, the metrics need to be much easier to access on mobile browsers.
Yes ofc it is centralised. The other solutions are PoW, which means you go back and try with your smartphone to outcompete servers, or payments, which they work but, you know, you have to pay for everything, ever for using a relay
Thanks for the feedback. It's very hard to make charts be great on mobiles, and frankly I am not interested in doing so. If one wants to do some analysis, then he can use a bigger screen.
Only the Real Odell β’is on TheForest π² .
If I'm not sure about someone with a large following count, I just try to broadcast their note to our relay.
wow i'm impressed by how incredibly fast it fetches data. Nicely done!
...it is down multiple times... only means you are still following messed up paradigms. when will you all #NamedDataNetworking #NDN be the network be the data be equal amongst equals stop connecting to servers and domain names start connecting to data instead. #when ?
Thank you! The secret is that it doesn't fetch the data every time the page reloads, but once every day. So the data is already there
yeah i assumed this had to be the case, still well done. The user experience is great.