Nico.cast🎩
@n
Working on a ML model which will output an 'address score' similar to how credit scores work (~low is 300-500, mid 500-650, high 650+) some interesting data: me: 529.52 @shoni.eth: 567.79 @cassie: 685.06 @jc: 698.19 @dwr.eth: 663.29 @dcposch.eth: 752.81
26 replies
0 recast
0 reaction
Nico.cast🎩
@n
Just if it's not clear: this pre-trained model will be open-sourced. I am training with FC addresses to start, and trying to get an idea on the scoring scale - are these too low? too high?
1 reply
0 recast
0 reaction
tldr (tim reilly)
@tldr
To answer, could you give a sense of what kind of thing the score is supposed to indicate? In credit scores, the purpose is to indicate something like future likelihood of promised payment. What is the analogous “thing measured” here?
2 replies
0 recast
0 reaction
🌶️ Zach Harris 🥀
@zachharris.eth
Seems like social capital?
1 reply
0 recast
0 reaction
Nico.cast🎩
@n
This is computed by a ML model which I will open-source - this should be the base reputation score of an address as it depends solely on public, on-chain data - any entities can provide their own scoring algorithm (e.g. Farcaster can give you base + 10 for being an active caster) final score = average of all scores
0 reply
0 recast
0 reaction
