Content
@
0 reply
0 recast
0 reaction
Ryan J. Shaw
@rjs
FID vs. OpenRank "Engagement" Ranking: - https://dune.com/queries/3901607 - *ridiculously heavily* skewed towards the earliest users of Farcast - Why? Because the algorithm starts off by selecting a trusted group of core users based on metrics heavily biased in favor of early users (see https://docs.openrank.com/integrations/farcaster/ranking-strategies-on-farcaster). - early users receive multiple boosts to engagement due to similar algorithms being used everywhere (e.g. Power Badge, Degen Tips), which results in a positive feedback loop. FID vs. Follower Count: - https://dune.com/queries/3901623/6557858 - nearly impossible for users who registered after a certain point to create the mega followings of the earliest users Why not use first cast or registration date? Dune's Neynar data doesn't have this <= 2023-09. If anybody has an alternative data source I can use on Dune, it would be much appreciated. Next: I investigate the degree of cliquiness in early users, if anybody is interested?
30 replies
14 recasts
70 reactions
0xLuo
@0xluo.eth
These two charts are excellent studies that indeed illustrate some issues. Regarding the number of followers, early adopters do experience the Matthew effect. But we know that a significant portion of their followers are bots that automatically follow them. We can also see that many people with a large number of followers do not receive much interaction. As for the OpenRank rankings, the charts clearly show that those who joined later (with larger FID) still have a significant chance of achieving higher OpenRank rankings. However, early adopters with high OpenRank are more prevalent and densely clustered. This is largely because most early adopters were invited by Dan, which effectively acted as a manual selection process. After permissless registration, the proportion of high-quality users naturally declined. Thanks for your work 🫡 3000 $DEGEN
1 reply
0 recast
2 reactions
Ryan J. Shaw
@rjs
Can't take the credit for the scatter plot approach... can't remember who it was, maybe @beachmfer.eth or @augusti or @apex777 who gave me the idea I think... You make some great points... I'm hoping to think of ways to differentiate between the explaining power of each possible effect
0 reply
0 recast
1 reaction
