Cooking the books -- the real story of whales and dilution.

The purpose of this post is to rebut certain claims about dilution made by @hisnameisolllie and others. You can read HNIO's take on dilution here: https://steemit.com/steem-stats/@hisnameisolllie/steem-distribution-revisited-top-steem-power-holders-to-distribute-up-to-34-3-of-their-control-over-the-next-12-months

Although i think centralization is counterproductive, and ultimately harms the credibility of the platform, that is outside the scope of what I am discussing here.

Good bad or indifferent, centralization is something that the platform, and its founders/leaders ought to be honest about. Because eventually, if "dilution" is promised by the founders (or by their unofficial spokespeople like HNIO) and fails to materialize, the platform will lose credibility. It will not only be perceived as centralized, but it will be percieved as both centralized and deceptive about it.

(* HNIO isnt the only one making these claims though he has gotten more exposure than most. If i seem to be going after him, its only becasue i believe hell take it in the spirit of honest debate )

HNIO's statistics make poor centralization metrics.

I intend to rebut @hisnameisollie 's post on its own terms, but i think it bears mentioning that the metric he uses represents a poor measure of centralization. At some later date, i might come up with a better method of measuring centralization in the system.

What HNIO does is list the stakeholders on steem in decending order of vest ownership, then culls off the top 1% as whales. The percent ownership of this cohort is then evaluated as a measure of centralization.

For a very simple example of why this metric does not work, consider the analysis of 7 day active users without steemit. We are told that in this category, the top 1 percent control about 83.16% of the voting power.

176,388M/212,101M=83.16%

However, this picture is incomplete. Although decentralization enthusiasts might have aproblem with any system where 1% of the participants control 83% of the voting power, all such systems are not created equal.

For example, a steemit where 83.16% of the voting power was controlled by the top 1%, and 16.84% was controlled by the next 1% and the remaining 98% controlled nothing. would be quite different from a system where 83.16 was controlled by the top 1% and the remaining 16% was divided linearly down the list.

This metric also disguises the real nature of a change in the 1% ownership rate. For example, if the one percent ownership rate went down from 83.16 percent to say 70%, many might view that as progress toward decentralization. But if that 13.16 percent all went to the second one percent, that is to say if it represented nothing but a transfer of power from the very very very rich to just the very very rich, most people would agree that this progress was only artifical progress toward decentralization. An edge effect if you wall.

The HNIO model analyzes the top 1% of accounts, not the top 1% of account owners

Because of this it severely underestimates the amount of centralization. For example, @dan and @dantheman are owned by the same person. Dan. SO for example, if the average top 1% user has 2 accounts, then his 1% statistic should have twice as many actual users in it.

But more importantly, evaulating control of the top X% of accounts, versus the top X% of account owners means that an apparent "dilution" of top 1% control could merely be a shifting of assets by SP owners who own top 1% accounts and also bottom 99% accounts. It certainly makes sense that top 1% accounts would have incentive to do this, if they felt that the public perception was that control of SP were too centralized. But even if we assume no active attempt at PR based asset shifting, its easy to imagine this happening just in the regular course of events.

Current SP distribution effects the distribution of new vests.

With the above criticisms in mind, i will take just HNIO's metric and evaluate it according to the real distribution of new vests.

The above analysis, as well as what follows apply to all the different models of activity discussed. I chose the 7 day acive users not including steemit because i think its the most accurate measure of regular account usage. And because I agree that generally steemit does not vote so should not be inculded in any measurements. However, the observations apply to any of the scenarios HNIO describes.

lets take the 7 day active users without steemit.

176,388M/212,101M=83.16%

Assuming a 343% increase in SP, HNIO correctly presumes that there will be a 34% increase in vests. Its actually a little less than that (because authors and curators are paid half with steem $) but lets assume everyone powers up their steem dollars. That means 72114 new vests will be created and given away

If none of the participation rewards go to the top 1%, at the end of the year our numbers would look like this.

176,388M/212,101+72,114M =
176,388 / 284,215 =62.06%

Note 1% ownership will go down from 83.16 percent to 62.06%. A 20% decrease.

But this isnt the whole story either.

Some of these 72114 newly minted vests will go to 1%ers. some will go to the bottom 99%

these newly minted vests will be distributed as such (these numbers are based on an arhag post). Im assuming all SBD will be powered up, so i am not differentiating how they are awarded (though in reality only half of curation and blogging rewards are paid in vests, the other half are paid in SD) Note also that for simplicity's sake, I have rounded down some decimals

52.4% will go for blogging rewards =39086mv
26.1% will go for curation rewards. = 18822mv
15% will go for block production =10816mv
7.5 will go to liquidity rewards. = 5408mv

Now, lets develop a model of how these rewards will be distributed.

liquidity rewards

Are currently suspended. No one gets them. When they were a thing, many/most of them went to abit, a top 1%er.

Because the precision with which you can set your price in the market place comes from SP level, its likely that this would be dominated by top 1%ers. im not going to count these for now.

blogging rewards

For blogging rewards lets assume that the top 1% get fair share (1%). This is absurdly, ridiculously false. But for just the sake of argument and becasue there is no built in mechanism to ensure that the top 1% get more of blogging rewards than the rest of us. I do, however, invite readers to look at the post history of top 1%ers to see how much they really get in blogging rewards.

176,388M+390M/212,101M+3908.6M+390M=
176799M/216400M

curation rewards

UNlike blogging rewards, curation rewards pay curators according to their SP balance. SO all other things being equal, the people who hold 83% of the SP should get 83% of the curation rewards. The bottom 99% who hold 17% of the sp should get 17% of the curation rewards.

For the sake of the model, and so that no one can say that I didn't give HNIO's theory every opportunity to prove out, I am going to assume that bottom 99ers are neraly 3x more effective as curators than top 1%ers, and give a 50/50 split.

176799+9411/216400+9411+9411
186210/

Block production rewards

go to the top 19 witnesses. There is no assumption i could make that would not have 100% of block production rewards go to the top 1%

186210+10816/235222+10816=
197026/246038=
80.07% for an decrease of about 3%

So even with no liquidity rewards, with 3X more effective curation by the bottom 99% than by the top 1% and with top 1% getting only 1% of blogging rewards, there is, in fact, only a modest decrease in SP concentration.