Making Steemit Better: A Proposal to Flatten the Rewards Curve

in #steem7 years ago (edited)

This is a joint post that is endorsed by each of the following people: @ats.david @clayop @donkeypong @gavvet @hanshotfirst @jesta @kevinwong @liberosist @sigmajin @smooth @snowflake @stellabelle @timcliff


Since Steemit began almost a year ago, there have been many complaints about the disparity in its voting power and rewards distribution. The main source of this discontent has been the n^2 system, which has concentrated voting power in the hands of large stakeholders, so called whales. In short, while there were good intentions behind this system and its lottery-style rewards, it has created an unfair game that only whales can play. The present system super-concentrates voting power for whales and leaves the masses with very little influence.

It is time to fix Steemit and settle on a better distribution of voting power and rewards. And that is why a large group of us (many of whom do not agree on much else!) have come together to support a more equal system. We are posting today to urge Steemit Inc. to adopt a near-linear rewards curve.

Criticisms of the Current n^2 System

“A handful of posts are draining the rewards pool”. “Whales vote for popular content, not the best content.” “Why do we need these large-scale curation guilds just to do the job of distributing whale rewards to authors?” “My votes count for so little”. If any of those criticisms sound familiar, the n^2 system is the heart of the problem.

There have been growing concerns that Steemit has become an uneven playing field because of this n^2 system (more technically, n^2 + 2*400*n, where n is MVESTS with 100% voting power). While this system discourages self-voting, it also makes the reward system more like a “handful of winners take almost all” lottery because higher value becomes much higher by being squared (e.g. 1:5 in raw scores will results in 1:25). More specifically, as more stakes are voted for a post, it gets much larger rewards as shown in a below figure as an upper blue line.

Image via @timcliff

While this system might be efficient to alleviate the self-voting issue, it also has side effects. There have been criticism complaining that Steemit is rigged in favor of larger stakeholders. For instance, roughly speaking, a user owning 100 shares may have 10,000 times greater power than those who has 1 share (in the real cases, it surely differ depending on voting dynamics). The nearly winner-takes-all system also pushes people and bots to vote for the most popular posts rather than selecting the best quality content. Moreover, the highly concentrated power has given to whales a considerable advantage, while many average users are left with little impact on distributing rewards. Consequently, Steem is being perceived as an unfair, whale-bot-oriented system rather than one that is human-oriented.

The community has been pushing for a change. Recent debates triggered by posts from @snowflake, @timcliff, @sigmajin and @ats.david provided some good suggestions to cope with these problems, but until yesterday, there was no consensus about which solution was most practical. Yesterday, after many hours of debate on, everyone signing this post agreed that Steemit must apply a flatten reward curve. Furthermore, we agreed to support the near-linear system outlined below.

Theoretically, purely linear rewards would be the fairest method (one has impacts either no more or no less than the degree to which one owns). However, there still remains the issue of self-voting that could potentially harm user participation. As suggested by @steemitblog today (post link), a newly committed reward distribution equation on the comment reward pool can catch two rabbits of fairness and anti-abusing, i.e. self-voting (be advised that this post only deals with reward distribution method, not dealing with the comment reward pool or removing curation reward). For convenience, we’ll call it modified n from now. The equation of modified n is n^2/(n+400)

Details of modified n

The below figure with two lines compares the n and modified n systems. The X axis represents MVESTS. As you can see, the modified n is linear-shaped with a small offset in the beginning. More specifically, where X is 10,000, the modified n has 9,615 (96.15%). This means that a post can utilize 96% of voting power if it gets 10,000 MVESTS amount (note that a top article with $130 had about 27,000 MVESTS).


Then how can it deal with self-voting? When we magnify the 0-1000 range (below), the graph seems more curved and has a significant difference from the purely linear n. That difference applies a discount in the beginning to discourage profits from self-voting. For instance, if I have 100 MVESTS (approximately $7,000) and vote for myself, my vote only has 20-MVEST equivalent power, which lost 80%. If I self-vote with 1,000 MVESTS, I still lose 28.5%. Therefore, one would need to get enough votes from many others to reduce the discount.


Example Case

How would this new rule change the landscape of reward distribution? As of yesterday, we simulated the new system with real data. Before reporting this, it should be noted that the existing votes are subject to the current rule, so real changes in the new system would vary somewhat from this simulation result. For example, “voting concentration” under the current rule will be less likely to happen in the new system since voting on highly-ranked content and lower-ranked content will return smaller difference in curation rewards.

Compared to the current rewards (blue line), the new reward distribution (orange line) shows a flattened curve with the range between $25 to $71, which is narrower than the current range between $19 to $132 (a dip in the graph is due to posts with high comment rewards). Out of 100 posts, 27 would have smaller rewards but the remaining 73 would have higher rewards.

In short, it would result in less clustering and a broader distribution of rewards.



We expect that the new system can bring about significant improvements in curation patterns and in users’ perception about Steemit, which we hope will fuel more demand for STEEM. We urge the developers to implement the new reward distribution method universally as soon as possible, and furthermore adjust a parameter (constant of 400 or 2E+12 in terms of rshares) in order to balance between two objectives: discouraging self voting and deflating the whale advantage. If more Steemians would like to register your agreement with this proposal to make the system fairer, please add your comments and votes to this post.


First things first, an enormous thank you to each of you: @ats.david @clayop @donkeypong @gavvet @hanshotfirst @jesta @kevinwong @liberosist @sigmajin @smooth @snowflake @stellabelle for spending what I can only guess to be a lot of hours and doing some serious soul-searching.
I will not pretend to understand the mathematics with which I think many of us are somewhat bamboozled. Any assistance in the levelling of the playing field has to be welcomed and I thoroughly hope you get the support of your peers.
I understood the breadth of posts affected in terms of reward range. It looks like a post with few votes becomes a more attractive curation exercise and one with many votes is less attractive.
If a post has 100 votes and a value of $0.02 and another has 40 votes and $25.00 potential payout, would the 100 vote post become that much more attractive? I apologise if that sounds like a daft question - just trying to get to grips with it!
At the same time, could you make minimum vote %, say, 25% - there are many who do not understand that minnows voting at 5% is sort of useless. Under this scenario it could become more counter-productive.
Thank you all again. Your work and ability to find consensus is very much appreciated!
Posting this in Facebook support groups too.

If a post has 100 votes and a value of $0.02 and another has 40 votes and $25.00 potential payout, would the 100 vote post become that much more attractive?

This proposal will narrow the gap significantly. If people have SP, their votes will be more powerful. That will remain. What would not remain is the enormous amplification that the system adds on top of the already-large disparity in SP holdings.

If you look at the third graph in the post, the $25 post would be somewhere toward the left and its reward using the revised formula would be much smaller. The $0.02 post would be somewhere toward the right and its reward would be much larger.

Many thanks for that explanation @smooth - you guys are all doing an amazing thing by finding consensus amongst yourselves first and by being prepared to explain and rationalise the reasoning and the effects. Thank you.
Have a great time for the rest of your weekend.

One clarification, as i think @smooth may have missed part of what the question was looking for.

In @ebryan 's example, the gap between the 100 vote, $.02 post and the 40 vote, $25 post will be narrowed, but the narrowing will be based solely on the total SP voting for the post.

That is to say that fact that the first post has 100 votes and the second post has 40 votes is irrelevant. The narrowing of the gap between a 40 vote post with $.02 and a 40 vote post with $25 will be exactly the same.

In a perfect world, i don't think rewarding the 100 vote post for getting more votes is a bad idea. But i don't think there is a pratical way that can be done without making the system exploitable by sibyls.

I mostly agree. I deliberately did not address the 100 vote vs 40 vote issue because also agree with your comment that it can't be changed (my proposal has always been to stop prominently displaying the vote count because it is misleading and making it more visible is not only confusing but creates an incentive to deliberately manipulate it, as we saw in the case of the Hot ranking before that was changed).

However, I do think it helps, in narrowing the disparity. A 40 vote post may (and will) certainly earn more than a 100 vote post, but even when that does happen the disparity would often be much smaller, in practice. Maybe that reduces the degree to which it seems unfair, maybe not, but it certainly can't hurt.

great approach for all Steemians I think given I can judge this yet @smooth

The tech part is not my strong point, so maybe someone else can answer you there. As for curation rewards, there are still some different views in the community over how best to handle them. But I think voting on posts that are not doing well yet (or from content creators who are not yet well known) will prove to be the better choice.

Thank you @donkeypong, I really appreciate all your efforts. I have already posted this in both the steemit and our Aspiring Whales and Dolphins FB pages. Great job - you have no idea! I think that this will go a long way to getting vote distribution where it should be - not a cure-all - one step at a time. Have a great rest of your weekend. Namaste.


Whatever works, you can count me in. Obviously looking for a solution means there is a problem, experimentation will eventually get us there.

This is by and large the biggest issue Steem currently faces. Good work @clayop.

Interesting. I'm all for giving it a go.

I stand behind this proposal. I see and know many people here who might benefit from a change as simple as this.

Agreed! I think about 99% of people would benefit... actually 100% if you have a long term view.

Well. We can't get anywhere if we don't look to see where we're going first. Always best to the look to the horizon and scan for problems on the way. Everyone seems to be doing that here.


Google Translate: I support it.

This is an outstanding post. The concrete examples and charts make it easier for the layman to understand. It would be amazing if this could be implemented!

I really do hope something like this can be implemented. As it stands I feel that Steemit is going to have a hard time really going anywhere at all. I would love to post content again on here but every time I think about sitting down and spending hours putting together a post and hope that a whale is going to upvote it just makes me cringe. So I just don't do it. It is a waste of time. Most others feel the same way. Getting upvotes here has nothing to do with quality. It is just this weird crap shoot. I just couldn't afford to spend my time here anymore and a lot of other people couldn't either. I could see myself being successful under this proposed model where as under the old model I won't be successful because I don't have any whale friends.

Here is the current Steemit model. The whales are the kings and everyone else is a jester trying to impress the kings.

Another thing to consider carefully: right now the shape of the curation reward curve is related to the author reward curve. Both are initially linear, and around the same point, the author rewards curve up and the curator rewards curve down. If the author reward curve becomes essentially linear, that will change the effect of additional votes on curation rewards. I haven't run the numbers, but I think it would make curation even less profitable for later voters.

So keep that in mind - making author rewards flatter without changing the curation curve will probably load curation rewards more towards early voters, which probably is a pro-bot change.

Great points. This is a first step and others probably will be needed. We were not able to find consensus yet on curation rewards, but hopefully that is something the community can discuss going forward.

Coin Marketplace

STEEM 0.28
TRX 0.12
JST 0.033
BTC 69746.86
ETH 3747.17
USDT 1.00
SBD 3.80