farcasterunion

I knew that spam accounts are created en masse, but it surprised me that some were very easy to filter out on Dune surprised me.

It gives an insight into the success rate of 'spam' accounts.
- 3,063 accounts that showed the same behaviour in groups of more than 15;
- There even seems to be a group of 500 to 800 accounts that have almost all received a spam label 0 after around 13 replies and 13 likes (disposable accounts);
- From these 3,063 accounts, there were 93 with a label 2, and 2 without a label. That's a success rate of 96.9%, the other 3.1% are false positives, and no labels.
- A large proportion of the accounts not identified as spam (89) were in smaller groups, which may suggest a different strategy of these groups of accounts. Here, the success rate is 90.4%. 9.4% of accounts slip through.

I've improved my script to detect spam more accurately and set some boundaries to prevent false positives (never 100% guarantee). It will also not detect all spam accounts, but accounts that show the same behaviour as a group.

The disposable accounts are not visible in this version due to choices to better map the clusters of accounts that are less often detected as spam.
https://dune.com/queries/5284227

For those who block spam accounts can use this query (below). If you enter your fid and run the script, you will only see the fids that you have not yet blocked. Always use your own judgement.
https://dune.com/queries/5269525

Interests: crypto, economics, TA & blockchain related innovations | 🔗 https://linktr.ee/cryptonjal 🔗 https://dune.com/nhejyht

HuntTown

@cryptonjal, you just received 6,000 applause from @pichi for being a based builder 👏

🏆 Today's Current Rank: 11th
👏 Claps to Next Rank: 139

Hunt Town Grant is a daily competition that selects the top three builders and awards them $hunt building NFT grants. Learn more at /hunt

Allowance will be reset in 09:08:28 ⌛️

Pi-chi.eth Artist /animeoutcasts
Hosting: /sakura /kyoto /momiji  /anime-manga /farcasterunion

https://portrait.so/pichi