Makes it pretty easy to see overall growth and what's happening in the network. Will post about follows cluster analysis later, but for now: how many unique people follow someone, per day?
(Database is available if you want a copy, see below)
Makes it pretty easy to see overall growth and what's happening in the network. Will post about follows cluster analysis later, but for now: how many unique people follow someone, per day?
(Database is available if you want a copy, see below)
Includes account details (did, handle, followers, following), follows, and PageRank calculations for each account.
Since no DMs here, DM me on Twitter: https://twitter.com/connerdelights
Includes account details (did, handle, followers, following), follows, and PageRank calculations for each account.
Since no DMs here, DM me on Twitter: https://twitter.com/connerdelights
Currently just using follows, but experimenting with likes now. Since it's calibrated to you, like spam isn't a problem.
Currently just using follows, but experimenting with likes now. Since it's calibrated to you, like spam isn't a problem.
Consider *just* people you follow and people they follow, build a PageRank trust graph. This is much smaller, and shows what accounts that your personal network trusts the most.
Here's my most trusted accounts.
Consider *just* people you follow and people they follow, build a PageRank trust graph. This is much smaller, and shows what accounts that your personal network trusts the most.
Here's my most trusted accounts.
Here's a graph of the raw values.
Clearly more followers → higher PageRank, but there are outliers. Most of the high followers / low page rank accounts follow spam, so have low-quality followers back
Here's a graph of the raw values.
Clearly more followers → higher PageRank, but there are outliers. Most of the high followers / low page rank accounts follow spam, so have low-quality followers back
Considering only accounts with more than 40 followers, here is the distribution of followers vs PageRank.
This gives "trust" recursively: if your followers are largely empty accounts → low score, if largely active → high score.
Explanation about what this means...
Considering only accounts with more than 40 followers, here is the distribution of followers vs PageRank.
This gives "trust" recursively: if your followers are largely empty accounts → low score, if largely active → high score.
Explanation about what this means...
Way more cozy!
Way more cozy!
Here's how many followers the entire Bluesky network has. Nearly every account gets 20-30 followers (follow spam), and then it drops off quickly.
If you have 100 followers, you're in the top 93%. The top 1% has 430 followers.
Here's how many followers the entire Bluesky network has. Nearly every account gets 20-30 followers (follow spam), and then it drops off quickly.
If you have 100 followers, you're in the top 93%. The top 1% has 430 followers.
Playing around with "who is the most underrated", I kind of like sorting by (followers rank - PageRank rank). e.g., you're the 100th most followed, and 50th highest PR.
This captures how far off your follower count is from PageRank.
Playing around with "who is the most underrated", I kind of like sorting by (followers rank - PageRank rank). e.g., you're the 100th most followed, and 50th highest PR.
This captures how far off your follower count is from PageRank.
From top 500 PageRanked accounts, these are the highest PankRank / follower counts ("most underrated")
From top 500 PageRanked accounts, these are the highest PankRank / follower counts ("most underrated")
(My crawl is a little incomplete, but this should be directionally accurate)
(My crawl is a little incomplete, but this should be directionally accurate)
You can make boba tea at home. Boba is on Amazon, cooks quite easily. Delicious.
You can make boba tea at home. Boba is on Amazon, cooks quite easily. Delicious.
(`pifft` is "People I Follow that Follow Them")
(`pifft` is "People I Follow that Follow Them")