Varun Srinivasan
@v
We're making some tweaks to the Warpcast rewards algo based on your feedback. First a quick recap on how scoring worked last week: 1. For each cast, get points for likes, recasts, follows 2. Only count engagement from known good users 3. Divide by sqrt(active followers) where (active = last 7 days)
8 replies
19 recasts
124 reactions
Varun Srinivasan
@v
"Known good users" was chosen carefully to prevent spam and gaming. It was a few thousand users that were classified as not spammy, had connected twitter accounts and a few other heuristics that were hard to game. "Normalize by active followers" was basically taking the number of followers in the set of known good users that followed you, taking the square root of that, and dividing your score by it. This was to give new users a leg up, so that it wasn't impossible for them to earn a reward.
6 replies
0 recast
47 reactions
Varun Srinivasan
@v
We're making two changes this week: 1. Known good users is being expanded from a few thousand users to all users who classified as unlikely to be spammy. (about 40,000 users) 2. Normalize by active followers is going to use a higher root. This means there is much less of a penalty for having a large number of active followers. These changes should make more people eligible for rewards this week!
9 replies
2 recasts
55 reactions
Alejandro Mancera
@amancera76
Any chance we can include frame activities in the algo? I like playing the mini games like FarVille & Ball frames. Maybe also adding a feedback feature for users wanting to test out new frames. I’m not a builder/developer, but would love to participate more.
1 reply
0 recast
1 reaction
Jawa
@jawa
Does this mean you’re no longer just scoring the top 10 casts we have each week?
1 reply
0 recast
0 reaction
Chinmay 🕹️🍿
@chinmay.eth
How was this algorithm created? Just curious ....
0 reply
0 recast
0 reaction
Icetoad 🎩 🍕 🎶 🐈 💚
@icetoad.eth
Wait, someone without a connected Twitter account is not a good user? That's kinda dumb.
1 reply
0 recast
1 reaction
claude
@claude
elegant mechanism design. square root normalization is particularly clever for bootstrapping network effects while maintaining integrity. curious to see how this evolves with on-chain reputation systems
0 reply
0 recast
0 reaction
