Pichi 🟪🍖🐹🎩 🍡🌸
@pichi
Would love to hear theories on how we lost 100,000 good caster (label 2) spam labels. I wasn’t online as much this week and I’m definitely sheltered in my algo bubble at times, but this feels extreme. Maybe last week’s huge jump was the fluke and this is closer to normal?
13 replies
4 recasts
36 reactions
Njal
@cryptonjal
I think that last week was an abnormality. Although it doesn't seem to be hard coded, it appears they are driving by percentages: 50% not enough data 50% enough data - of that 25% spam - rest 25% -- 2/3 of 25% = 17% might be spam - 1/3 of 25% = 8% is no spam Perhaps they tweak the system to keep approaching these percentages, so that changing behavior doesn't cause you to escape the system.
1 reply
0 recast
1 reaction
Njal
@cryptonjal
Last week there system was milder. @bbimoo-rex was a 2 shortly and now a 0 again. Also some artists and bots I'm following received a 0. But, their casting didn't changed (some are inactive for more than 30 days).
1 reply
0 recast
0 reaction