Today, the devs commited two changes. First one is renewed reward calculation based on linearity. And another is applying this reward curve to all reward system. This is great news because many of us have been requesting more flattened curve for fairness. I first appreciate the devs' hard work and their passions for Steem.
Details of Change
Technically, the new curve is calculated as
(r^2 + 2*a*r)/(r + 40*a).
r denotes sum of Steem Power denominated in
MVESTS that is given to a post, and
a is constant
55. Seems quite difficult... but I can show how it works.
The graphs has X axis of received MVESTS and Y axis of effective MVESTS. The bottom graph is a zoom-in of 0~5000 MVESTS range. In the perfect linear system, if I cast 1,000 MVEST, my vote is equavalent as 1,000 MVESTS reward shares (Currently, 10,000 MVESTS post gets around $20). But in the modified linear system, 1,000 MVESTS end up with 344 MVESTS because of anti-abusing (self-voting or collusive voting) discount. Then how much voting power is needed to have less penalty?
The below graph with four lines presents voting efficiecy at given MVESTS. Perfect linear system always has 100% because 1 MVESTS always effects as 1 MVESTS. Amont them, Currently commited setting is blue bold line, which has denominator of
(r + 40*a). Green line uses 10 instead of 40, and light green uses 20. Meanwhile red uses 80. These numbers change voting efficiency along with MVESTS, and higher values mean stronger anti-abusing penalties.
With the commited value of 40, if a post got voted by 4,600 MVESTS, the next 1 MV vote has about 0.9 MV power. But with the value of 80, 9,400 MV is required. Meanwhile, with the value of 10, a post only needs 1,000 MV to give +90% efficiency to following voters. The below table shows required MV to reach 80%, 90%, 95%, and 99% efficiency.
|10*a||560 MV||1015 MV||1660 MV||4300 MV|
|20*a||1250 MV||2200 MV||3585 MV||9180 MV|
|40*a||2620 MV||4615 MV||7420 MV||18920 MV|
|80*a||5360 MV||9410 MV||15090 MV||40000 MV|
A higher anti-abusing power sounds good. But it has pros and cons.
- Higher anti-abusing power discourages self-voting and collusive voting.
- Rewards are relatively more concentrated to high rank posts
- Some mega-whales will have advantages to put a post beyond penalty box. E.g. One vote from 6,000 MV whale can make a post have +80% efficiency, while 1,000 MV whale can make it +35%.
- Vote concentration, mainly by bots, still can exist. High bar implies there are fewer posts having good efficiency (e.g. +95%). Then, for-profit bots are able to choose targets more easily, maybe there will be about a hundred candidates. But if hundreds of posts have +90% efficiency, bots will have harder time to choose and they may need better and complex algorithm.
Overall, I am very glad to have more fair reward system universally. But I am still suspicious that this will reduce vote concentration problem meaningfully. A virtue of linear-based system is that any abuses cannot have power more than he/she owns. That is their vote efficiency is always below 1. I think linear system already significantly resolved self or collusive voting issue. Now is the time for fairness. Give many dolphins more power. Make bots have hard time. Give more rewards to more authors. And we will see Completely New Steemit!