Data

Don't remember if I shared: Publicized my logic to ~approximate duplicate/spam activity (input to ~likely sybils)

Why open it up? Encourage more open models, so we can progress & stop people from using "active addresses"

Still messy, but you can see ideas: https://dune.com/queries/3783798

NOTE: Not for production

data @ op labs, building @optimism 🔴✨ | go /gators 🐊  | casts are my own things | msilb7.com

Basic Logic:
- Find duplicate tx histories (method id, to addrs, time elapsed, etc)
- An address with few txs is "too early"

Note: This logic is bad, has MANY false positives & negatives. This is far from sufficient to be used in "production" cases. It's v easy to go undetected

Where is this used:

1. Try to segment onchain activity by type of address (also incl. high-frequency bots)

2. Superchain Health dashboard to ~try to get closer to the count of onchain actors (but still v far off - recent lift in Aug-Sep is likely false): https://docs.google.com/spreadsheets/d/1f-uIW_PzlGQ_XFAmsf9FYiUf0N9l_nePwDVrw0D5MXY/edit?gid=584971628#gid=584971628

How can you help:

1. See where this logic does/doesn't make sense
2. Build something better & share it with the world

I feel like the best short-term approach is some kind of hurricane spaghetti model, where we aggregate a bunch of plausible ideas (identity/trust, patterns)