pull down to refresh

The algorithm for rewards has been updated to distribute rewards more uniformly. It went live last night.
The number of stackers rewarded last night was 5x the number of stackers rewarded prior to the change.
The algorithm still only rewards zapping, a change we initiated as an experiment, but we plan to run other experiments after observing the impact of rewarding only zapping with more uniform distribution.
We'll also be adding more sats to the rewards pool next week starting monday as we run an advertising experiment and begin to distribute the 18m recovered crazy rewards bug sats (aka the forbidden sats).

details

The reward algorithm takes the top 21% of posts and top 21% of comments and rewards the zappers of that content. This part hasn't changed.
Prior to this change, a stackers reward portion were calculated as (sum((1/(zap_order))*zap_amount*relative_item_favorability*zap_trust))/sum(numerator across all stackers).
Where zap_order is the order in which the stacker zapped the item, i.e. 1 for 1st, 2 for 2nd and so on. This term means we reward early zappers more.
zap_trust is a score from 0.0-1.0 and depends on how much other stackers agree with your zapping historically (on some level it represents a prediction of the probability that other stackers will agree with your zapping).
@kr noticed that these reward proportions were heavily skewed. We were only rewarding ~15% of zappers every day.
This release changed a stackers reward proportion to: (sum((1/(ln(zap_order+e))*log(zap_amount)*relative_item_favorability*(ceiling(zap_trust*3)+1))/sum(numerator across all stackers).
1/(ln(zap_order+e)) - this term change means you'll still be rewarded more for zapping early, but no longer disqualifies you for rewards if you aren't super early.
log(zap_amount) - this term change means you'll still be rewarded more for zapping more, but no longer disqualifies you for rewards if you zap slightly less.
(ceiling(zap_trust*3)+1) - this term change means you'll still be rewarded more when trusted, but no longer disqualifies you for rewards if you are low trust or untrusted.
We also significantly handicapped all accounts associated with SN as they are natural outliers.

bonus

  1. editing your nym is now debounced (issure discovered by @ekzyis and fixed by @WeAreAllSatoshi)

btw

I'm paying sats to everyone mentioned in these releases for their open source contributions. I'll pay even if you don't ask (no one has asked), but if you want to be paid a specific amount for something, please say so and we'll work with you to get it shipped.
I'm still getting more in early-tipping zap rewards than I'm zapping I think, so nice work.
Are zap rewards a security? Should I register these with the sec?
reply
Uncle Gary did mention you at dinner the other night.
all I know is if it’s called Mutiny they’re probably up to no good
reply
Sounds like an endorsement I should put up on our marketing page
reply
Call it family wallet. They wont suspect a thing. But also family is the direction we need to go. imo. sorry for my very unsolicited and cheap opinion
reply
Confirmed, I have been paid in sats for my contributions 😊
ETA: I also received rewards last night for the first time in a long time, so that was nice!
reply
Same here. Nice work guys!
reply
I was pleasantly surprised this morning when my reaction to my rewards went from "what, only?" to "wait, what did I zap that earned me so many sats?"
reply
I know. We record this info. We just need to display it. Markets are more efficient when all the info is available.
If I can get the work I have planned today done, I'll try to working on displaying this info tomorrow.
reply
Keep up the good work.
reply
Great work
reply
The circular economy maturing!
reply
Thanks for explaining the code to us. I bookmarked this for future consumption (I.e. when I get more familiar with coding) haha
reply
What about the user behind top post (not a zapper) ?
reply
Once this experiment is over, we'll be reintroducing the other types of rewards in some fashion.
reply