I have read many people saying that they would like to see a numerical analysis of the forthcoming EIP. I have seen nothing but promises regarding the change from 25% to 50% curation rewards, so here is my mathematical analysis. If I've missed any similar posts, then please link to them in a comment; thanks.
What is Curation?
Before looking at the mathematics, I wish to spend a little time looking at some psychology. One thing that causes endless arguments is Steem's abuse of the word "curation". This Abuse of Language Syndrome has infected all discussion on this issue. The virus was first injected by Steem's original white paper, which claims that all upvotes are curation. Really!? Does anybody believe this? Most recently, steemitblog's own post states that, "Under these changes self-voting and bidbot usage will become less profitable than curating good content." Again, does anybody believe this?
So let's try and clean up the language first. Curation-reward-bots are not the same as human-quality-curators. Is that not obvious? Algorithmic-curation is very different to quality-curation and, by their very difference, bot-votes will always be more profitable than manual-votes.
Algorithmic-curation takes place largely within 15 minutes of posting - apart from those sniping upvote-buyers - and its only aim is to maximize curation rewards.
Quality-curation is done by humans and often many hours after the post was made. This in itself severely limits the curation rewards that such human-curators can earn just from their own upvote. Although curation groups do give their curators some incentives for their work, the issue here is that so-called curation-rewards form a very small part of the income of both the individual and the group because, again, their quality-votes take place long after the bot-votes.
One advice I often give to curators is to add their favourite authors to their own auto-voting-bot so that they too can earn longer term from algorithmic-curation.
So what words can we use so that people are not talking at cross-purposes about two very different modes of upvoting? I would suggest the simple human-curation and bot-curation. Pick your own favourite, but using "curation" when two people have different things in mind causes friction and is counter-productive.
With that out of the way, here's some mathematics.
The Upvote Trading Model at 25% Curation Rewards
I'm going to make some assumptions as we go along. Those assumptions are based on looking at the numbers on the blockchain for 2 years. Such assumptions are not universal but are very good estimates of the averages I've seen for various parameters and metrics. If you wish to quibble with some assumptions, then also supply your own calculations all the way through the process.
The Upvote Trading Model as currently exists on the Steem blockchain is a two-party exchange that in its essence works like this: User A sends a transfer to User B, then User B upvotes User A. That's it!
This is the source of so many profits and is also the reason why algorithmic earnings are taking over the blockchain. Before I move on, I must also stress that there are some posters who are well rewarded by the community and that such rewards are way beyond what they could earn algorithmically. There are also some whose posts are being rewarded for other work they may do behind the scenes as devs or mods etc. Thus, posting good content and being rewarded for it can work, but the problem is that posting crap, or even not posting at all, can also be highly rewarding.
OK, let's assume user A sends user B a transfer of 1 STEEM. I would also like to assume an ROI of 0% for A; that would mean an upvote from B of about 1.333 STEEM and makes the next numbers ugly to look at. Let's instead assume that A was lucky and gets a 5% ROI on the transfer. This results in an upvote from B worth 1.4 STEEM.
Now, 25% of that goes to curation, leaving A with rewards of 1.05 STEEM. Now, B already has 1 STEEM from A and will later also receive some "curation rewards"; these are likely to be far less than 25% and I shall estimate 15%; then 15% of the 1.4 STEEM upvote comes to about 0.2 STEEM. This gives B a total income of 1.2 STEEM.
So, let's firstly sum up what has just happened:
A has paid 1 STEEM for 1.05 STEEM worth of author rewards, plus let's not forget some curation rewards too, assuming A has strategically self-voted.
B has gained about 1.2 STEEM for a 1.4 STEEM upvote.
User A has a return of 105% and user B has a return of 86% based on the value of an upvote.
Why do people delegate to vote-selling services? Well, 86% is significantly higher than a self-vote that may only yield the 75% author-rewards.
Bidbots are just a particular kind of vote-seller and are able to increase that 86% return by allowing each round of bidding to end up with a negative ROI for the bidders.
Also note that as the Steem blockchain is currently yielding about 21% APR for an upvote, that vote-sellers can thus generate about 18% APR - even by being generous!
One last important thing to note is that the vote-sellers and their delegators are getting 11% more than the 75% expected as an author, and this comes to about a 15% increase above the author-reward rate. This figure will become important later.
The Upvote Trading Model at 50% Curation Rewards
Let's see what happens if we switch to 50% curation-rewards and 50% author-rewards.
We need to keep something constant, else the numbers become hard to compare, so I'm going to assume the generous vote-seller B gives the buyer A an effective reward of 1.05 STEEM on a 1 STEEM transfer.
At 50%, this means B must issue an upvote worth 2.1 STEEM so that half will go to A and the other 1.05 STEEM will go to curation-rewards.
Now, curation-rewards have doubled, but the curation algorithm is not being changed. That curation-rewards algorithm depends on the relative upvote size of each new upvote compared to the existing upvotes. It is thus not very clear whether my estimate of 15% curation to B will hold, or may even increase. The big assumption here is that although the rshares allocated to curation-rewards has doubled, this does not mean that all upvoters will change their own voting amounts. I've done some mathematical analysis of the formula and estimate that it is likely, because of the doubling of these bought upvotes, that they may make slightly more in curation; my best estimate is 20%. Let's look at both numbers to see how sensitive the calculations are.
With 15% curation reward income
Anyway, with 15% curation-rewards on a 2.1 STEEM upvote means about 0.32 STEEM will come back to B, giving the vote-seller a total income of 1.32 STEEM from a 2.1 STEEM upvote.
So, what has just happened?
User A is still getting 1.05 STEEM from a 1 STEEM transfer, whereas B now has 1.32 STEEM from a 2.1 STEEM upvote.
The return for A remains 105% whereas that for B has fallen to about 63%.
On the face of it, this is a large 23% drop from the 86% return with 25%-curation.
But remember that author-rewards have also now dropped to 50%, so that a 63% return is +13% on a self-vote; this means that the profit is 26% higher when compared to self-voted author-rewards.
This is a larger differential than now!
With 20% curation reward income
Replacing the 15% curation-rewards with 20% means that vote-seller B receives 0.42 STEEM in rewards plus the 1 STEEM transfer, giving a total of 1.42 STEEM from a 2.1 STEEM upvote.
This gives a return of 68% based on the upvote and +18% based on the 50% author-rewards, or 36% higher than self-voting!
What just happened?
The idea of a profit really depends on the basis for the calculation. For the vote-buyer this is obvious as they are sending a transfer of STEEM and wish to see that used profitably. For the vote-sellers, and their delegators, profit is relative to other forms of income on the blockchain and can either be measured based on a full upvote or on the author-rewards of a self-vote. Many delegators are just looking for a passive income without the complications of posting crap and upvoting it ten times a day - that isn't exactly passive. The leasing market tends to be priced just above the returns on author-rewards and below the APR of a full upvote, so these are valid calculations.
My calculations above have shown that income for the vote-sellers will drop significantly but that may not be so important for the delegators as their passive returns as a percentage of author-rewards will actually increase!
This means that vote-sellers such as bidbots may need to do little else but adjust their ROI and algo to mitigate some of their downturn in profits; other vote-sellers may need to adjust expectations.
One last thing to note is that vote-buyers are often in profit from other upvotes that are only there because they are chasing curation-rewards in the knowledge that a large upvote will later arrive. Such bidders do not require the bidbots themselves to have a positive ROI because of all the other votes they attract.
There are also vote-buyers who just do not have enough SP to even think about self-voting as an alternative. Again, any small profit comes largely from bidbot sniping algorithms, so they can tolerate a small negative ROI from their bids.
In Conclusion
We shall see how the whole ecosystem reacts to HF21 and then see how the macroeconomics of Steem reaches a new equilibrium. After HF19 there was a large drop in the reward pool that took a number of weeks to reach an equilibrium. In contrast, HF20 had no effect on the reward pool, save for the technical blip during the first week and that took a couple more weeks to resolve itself.
Let me return to the top: "Under these changes self-voting and bidbot usage will become less profitable than curating good content." They should have stopped at "... less profitable." As we have seen, self-voting will be less profitable as rewards drop from 75% of an upvote to 50%. Bidbots and other vote-selling services may only need to implement small tweaks. This is because the delegators to such services may well be earning more than before compared to the lower author-rewards. This is really the flaw in this whole system. Existing delegators will, indeed, experience some drop in the net earnings they have come to expect, but the increased premium above the lower author-rewards may well encourage new delegators to join.
The human curators will always be voting hours after a post is made and their upvote may well yield under 10% in curation-rewards - it will be double of what it was before, but just a larger pittance. This may, however, be good for those curation groups that deliver large votes; hopefully this will feed back to their hard-working individual curators. For genuine bloggers, there are a growing number of sites opening up such as PALnet. As they are all linked to the Steem blockchain, such authors may well be earning multiple tokens and coins from each post. This is going to be the way forward in the medium term.
If I may quote myself from an article posted some six months ago here: "The current set of encoded rules do not, and will not, by themselves, promote the kinds of social interactions people expect." HF21, or at least the EIP part of it, will be no exception. And also, from the same article: "It can be argued that Steem will be fine as a source of high interest profits for all the banker-members; that the social fabric can be built on top of this financial machine and that social-bankers can help fund this development."
In conclusion, let's sum up the headlines: the 50-50 split in author/curation rewards will see lower rewards for authors (obviously); lower net returns for vote-sellers and their delegators but higher percentage returns due to the lower benchmark of author-rewards (hence no disincentive to continue with just minor adjustments); vote-buying will remain potentially profitable due to curation-seeking upvotes sniping the vote-sellers; and higher curation-rewards to human-curator but only significantly so for large voting accounts. This in no way makes human-quality-curation more profitable than algorithmic-curation, as is the claim by EIP and HF21.
Is this what you were expecting?
So... where will all those extra curation-rewards go?
Algorithmic curation-reward mining.
Expect to see more tokens and projects "sharing rewards".
Be prepared.
As a coda, I believe there are possible solutions so that the economic code within the Steem blockchain can serve its original mission to be a social media blockchain and not just social media on a blockchain. The level 2 dapps such as PALnet, and others, have become a necessary development but they are also an opportunity that gives us the time to look back at the original economic code; keep the parts that work, such as the macroeconomics of STEEM and SBD but investigate those parts that seriously need an update, such as the microeconomics that everyone experiences.
I propose to submit a working group to SPS - anybody with the skills and interest may express their wish to join this project. Thanks for reading.
images: pixabay (edited)