You are viewing a single comment's thread from:
RE: Whitepaper Discussion on Voting Abuse
nˆ2 was gave too much power to a few, while linear may indeed not be the right approach as well. I read from many something in between is the best approach. Do you know if something inbetween is possible to implement technically? I think that was part of the discussion, the inbetween functions was not possible to implement, or was/is that more a BS argument?
The "correct" formula, if we assume that "value" is commensurate with network "connections", is n.logn. calculating logs is computationally a pain, but can easily be approximated by n.(number of digits of n)
But anyway, raising the influence at the top end is just for rich accounts. I proposed some time ago how to "squash" the effect of excessive voting for one account, whether self or other. Just think what n^2 actually means: an account with 100k SP compared with 1000 SP will have 10,000 times the voting power.
The idea was/is that the few rich accounts are less bad for the community, and this idea is enhanced since all the more 'bad' seen in recent days with so many new accounts here at Steemit. However, the question is if that is indeed the case in relative terms. I have no stats on this. I must say, before we changed to linear, some of my posts got way more rewards, then I now get. I suppose this has to do with the linear curve where the powerful voters are not that powerful anymore; But it could also have to do with other dynamics that changed. Who knows.
I realise I'm 2 months behind here but ...
If the curve was n.log.n then how much larger would 100k SP be compared to 1000 SP?
It depends a lot on the constant factors. Using log_2 and no fixed constant an example:
SP = 1 weight = 0
SP = 2 weight = 2
SP = 4 weight = 8
SP = 8 weight = 24
SP = 16 weight = 64
...
SP = 1024 weight = 10240
SP = 131072 weight = 1703936
So as you can see here doubling the SP results in "a bit more" than double the weight and increasing SP by 128x increases weight by 166x (compared to n^2 where doubling the SP results in 4x weight and 100x the SP results in 10000x the weight).
There are some technical reasons (some of which I understand and some of which I don't) that complicate implementing n log n. There are also philosophical/fairness/social reasons to want to preserve linear and not convey a programmatic advantage to the largest stakeholders in excess of their stake (even if only by a relatively modest degree).