Overhaul of Curation Rewards
With July 4th rapidly approaching we are running out of time to review and tweak how curation rewards get applied. One of the reasons we opted to wait until July 4th before paying out rewards was to give us time to evaluate and tweak the algorithm based upon actual usage patterns.
Lessons Learned
Over the past month and a half we have learned quite a bit about user behavior and interaction with Steem. Here is what we have learned:
Curation Rewards don’t help the little guy
The exiting curation reward allocation strategy heavily biases rewards toward those with a large stake. This creates a centralizing effect that doesn’t sit well with many people and ultimately defeats the goal of making it easy to recognize participation.
Bandwagon Effect
The existing algorithm encourages people to pile on a single post. This distorts voting behavior and ultimately could result in unintended / biased curation results.
Big Bad Bots
There are bots that get the jump on humans. While many people correctly point out that bots add value, there is a very real feeling that they suck up a disproportional amount of curation rewards. This combined with bandwagon effect could evolve into undesirable behavior.
Cognitive Load
People face a cognitive load trying to figure out how and when to vote. This makes voting stressful and causes people to vote less than they otherwise would. We would like to reward honest people voting honestly without making people play a game. We are successful if honest behavior is rewarded more than attempted abuse. We fail if people spend more time thinking about their rewards than they do actually evaluating content and expressing their honest opinion.
Comments get little Love
One of our original goals was encourage quality discussion. The algorithm as it exists now discourages people from voting on comments. This happens because it dilutes their voting power which in turn reduces their potential curation rewards. Those posting comments get significantly less votes than those posting top level articles which means they often get next to nothing for their effort.
A new Approach
Our team has been discussing these problems and potential solutions at length. Many on Steemit have also proposed interesting ideas. Ultimately we concluded curation rewards would have to take a radically different approach if we wanted to maintain a more egalitarian platform. So we started asking ourselves what kind of behavior do we really want to see. This is what we came up with.
Vote or Post Every Day
Users who login and do something every day remain engaged. This engagement will naturally lead them to discover new content they will want to vote on. We don’t care what they vote on, so long as they show up every day.
This can obviously be automated by bots, but it has the important quality that bots have no advantage over regular people. The network already rewards “bots” who produce blocks and mine so giving them a small extra bit for voting daily doesn’t really change much. It is more important who isn’t getting the rewards: those who do nothing.
Discuss Quality Content
If a post is going to earn $4000 of rewards, then the community had better discuss it throughly. Any content that isn’t worth discussing is probably not worth rewarding. It is the discussions that ultimately provide ‘reviews’ of long articles. Discussions are where fact checking occurs. Plagiarism is also identified in the discussion. You could easily argue that the discussion is a better kind of curation than voting. Discussion helps people decide what to vote on and is ultimately something that bots cannot do.
More importantly, it is far easier for most people to post a quick comment than to write an article that becomes popular. This creates more opportunities for average people to be recognized on Steem.
Producing Quality Content
It has become clear that Steem is evolving toward rewarding original content over rewarding cross-posting content. We need to maximize the incentives for original content creators above all.
Proposed Allocation of Curation Rewards
Here is the new algorithm that we would like to use to divide / allocate the curation rewards:
Activity Rewards
30% of existing curation rewards will be redirected toward activity rewards. This represents a total of 15% of all (content+curation) rewards. To maximize your activity rewards you must vote, post, or comment at least once every 24 hours. These rewards are distributed proportional to the amount of Steem Power your account has. There is no advantage to being more active other than the joy of reading and rewarding your favorite contributors.
To minimize abuse and/or delegation to bots, only accounts with simple posting authorities consisting of a single key can qualify for activity rewards. If you attempt to use advanced multi-sig configuration on your posting authority then you will not quality for activity rewards.
Comment Rewards
50% of existing curation rewards on top-level posts will be redirected toward comments on that post. This means if you see a post with a pending payout of $10,000 that $2,500 will be allocated toward comments on that post. Comments at any level will divide up the $2,500 proportional to the total amount of Steem Power voting on the comment.
Removing Parent Rewards and increasing Author Rewards
We originally intended to reward the parent for stimulating discussion by sharing part of each comment’s reward with their parent. This process has been removed because it contradicts the direction of payments that we now have for comment rewards. To make up for lost parent rewards, the remaining 20% of existing curation rewards will be given to the author.
Summary
Steem currently allocates 2 STEEM every block toward content + curation rewards, after these change the 2 STEEM will be divided like so:
60% to author of post
25% to comments (at all levels) on original posts
15% activity rewards
Conclusion
Under the new rules it should be possible for more people to earn Steem Power just by being involved each day and adding to the discussion. Content creators will have an active and engaged audience who will likely share what they are discussing with others via other social media platforms. We will have greater ‘curation’ as people post comments that review / critique / add to the original content produced.
People will have financial incentive to be first to reply to potentially popular content with a comment that itself will likely receive many votes. This process makes the people voting a turing test for curation and minimized the potential profits of voting bots.
Overall we feel that these changes should maximize the community participation.
I respectfully disagree on this. How do you envision a discussion on articles that merely describe procedures in a how-to fashion?
You won't achieve what you want with this setup but merely kill businesses such as Streemian.com that NEEDS this feature for onboarding people that are using RSS-capable blogs. If this change goes through I can stop working on Streemian.com right away because I can't make any money out of the people I bring into Steem. This is really bad!
Let me quickly describe to you how the profit model of streemian.com works:
This means, the more people, we bring in that produce good content, the more profit they make and the more profit we can make by upvoting our content.
This however requires multisig hierarchies so that the original poster can use his account in full.
On the other side, you want to prevent bots from voting. Well, I must tell you that won't work because I can replace compromised posting accounts anytime. That means I simply share the same private key with my bot and I will still get rewarded.
Hence, this needs reconsideration!
I'd also appreciate a re-look at this particular aspect of the change - it directly affects the work we're doing to develop services to remove friction for steemit adoption - @dan / everyone, can we get your input into this - can you think of anyway we can achieve the same outcome from both a streemian and delegated bot perspective - whitelisting of delegated authorities for example. I imagine this capability we're using with streemian will also be of use to other services that want to build on steemit.
I really think this changes will promote the use of steemit, posts will have more comments and replies and this is a good thing. Well done!
Question: Will this change the current curation payout for the 4th of posts already created or will it apply to all posts from now on?
Yes, this will change the current curation rewards which we have maintained as "pending" until July 4th. This means that authors of posts will likely see slightly larger payouts, active minnows will likely see larger payouts, and commenters will see much larger payouts. The people who will see smaller payouts will be the whales that have upvoted popular content.
In other words, this change prevents the "rich from getting much richer" due to the n^3 nature of curation rewards. It will also create incentive for everyone to go back and review / comment on all existing posts prior to July 4th to make sure they are reasonable.
BUT only commenter’s that have comments with votes! And that is not the case for the majority of comments because nobody had the motivation to vote on comments... Explain me why they will vote comments now? Are the voting power not reduced again after voting on comments? What am I missing?
PS each comment from now on should de-facto get a minimum auto-self-upvote (for example 1% of our voting power, the exact % can be discussed further) that way SPAM RESPONSES should be less, and potentially REPLY-BOTS will get weaker !
Wow!! this is amazing for the little guys like me who participate a lot.. Thanks!
Awesome, I just can't wait for people to see what you're doing here, this is so exciting, thanks.
Great change. Something tells me this post is going to end up with a ton of comments!
So what will the voters get? Nothing? So why bother voting for others? I guess the bots will post rubbish replies on every [good] post and vote for themselves, normal users will have no chance to down-vote them all.
Voters get content catered to their voting preferences. In exchange for voting every day they get a fixed rate of return on their steem power. In most cases small voters will get something where as before they got nothing.
Rubbish replies should be actively downvoted which will earn the bots nothing. I think we will have to disable comment rewards for any comment posted 2 hours before payout just like we disable the ability to change your vote before payout.
All payouts are on the basis of n^2 which means someone who only votes for themselves will be earning "below average" rate of return compared to someone who gets the benefit of n^2 when many people vote together. To overcome the power of N^2 would require a massive amount of spam voting which would dilute voting power tremendously cause you to fall further behind the power curve.
Previously voters were discouraged to downvote. Has this changed? If one person takes it upon themselves to downvote a bunch of rubbish comments, how will they be affected for doing a good deed? What if we had an option to "flag" a comment as inappropriate and if enough flags were clicked the post was automatically downvoted by the Steem system instead of the users voting power being affected by the downvote? Although I'm sure a system like that could be abused.
Why are we discouraged to downvote? Crappy content should be voted against.
It's a result of current rules, but not intended. One of the issues need to be resolved.
I agree downvote have to be rewarded somehow. The flag system you have proposed would need a anti-abuse system also. If not bot would flag everything that isn't theirs.
Please correct me where i'm wrong
" Any content that isn’t worth discussing is probably not worth rewarding "
This sounds to me that we urge you to write or share posts that is dear to the current community and we discourage you to write anything else. Just write how you like steemit and introduce yourself ( how was your life so far as we are something like a reality show). Try as hard as you can to not guide thematology.
" It has become clear that Steem is evolving toward rewarding original content over rewarding cross-posting content. We need to maximize the incentives for original content creators above all. "
This sounds to me that we don't like variation. When you look at the front page of any social media think what you are doing. You scroll down, check different topics, you pick what interest you and you read it. Not only because you want to learn something ,or motivate yourself but to get news, have fun, pass your time. By doing this, there is no point for someone to post news or something funny or anything else like kittens :) because he or she will get a 0. No motivation there. You get traffic by cross-posting content and that's what we need. Everyone is trying to write his unique or not article and thumbs up for doing it, but we lack of a powerful feed. And here is where i ask, what are we trying to accomplish? Do we want this place to be an e-magazine with a few good writers and people who just read this place or an interactive social media, in which everyone posts anything and people talk about, agree or disagree ,fight or make friends
" Activity Rewards
These rewards are distributed proportional to the amount of Steem Power your account has"
" Comment Rewards
Comments at any level will divide up the $2,500 proportional to the total amount of Steem Power voting on the comment."
My opinion is Steem Power to get something indeed this is the core of everything but not have a huge impact
Excuse me for the long comment
I agree, this was the only thing I have issues with in the changes (majorly anyways). I think "original content" that adds value should be rewarded more then say a copy paste article. But by all means discouraging copy and pasting or saying it may not bring the same value as an original post is rubbish. At one time it was an original post and if it has just been released that day then I see no reason why a brand new article that wasn't posted by the original author can't be taken and posted here by someone else earning them financial compensation.
I wanna see this beautiful thing grow, it can cover multiple topics, people are saying it won't be the next Facebook and I see no reason why it can't be. I see it as a mix between Reddit and Facebook (potentially) taking the best of both worlds and mixing them together. I see no reasons why we can't have people from around the globe all discussing a variety of topics.
Crypto's may have been the first group interested in this but I don't see why it can't "evolve" where we have entire groups who enjoy meme's of cats
;-) lol.
So curation/activity rewards are maximized after a single vote per day?
If I have nothing to add in comment, why would I bother to read or vote for anything else?
Does it matter how popular what I've voted for becomes or is my activity reward the same regardless?
It seems like you're eliminating the incentive to be choosy about what you upvote and you're relying on altruism to sort the best content. Everyone should just vote for their own comments everywhere. If there's no penalty for picking the wrong horse, I'm going to pick myself every time and never vote for anyone else, to maximize my rewards with least effort. I hope I'm misunderstanding something but this seems to remove a vital component and potentially incentivize half-hearted spam comments.
There is still the n^2 rule which means that voting for yourself will cause you to earn below-average returns compared to voting for others. Writing a bot to "mine" for you by posting and voting on yourself is unlikely to be profitable because others will write counter-bots that negate the profits of anyone found attempting to milk the system in such a way.
To me the new rewards system isn't clearly described in the post. I don't see where voting is rewarded at all, aside from the active-once-per-day distribution. I guess what is not clear is what elements of the old mechanism are retained or modified.
Would you mind summarizing the new rewards mechanism without reference to the old one?
The problem is : by voting for others you will gain nothing at all, if nobody else votes for your comment. So how could voting for oneself be below-average?
I doubt there will be much 'counter-bots' which just be written to punish others. Haven't seen much incentive on that.
The incentive to vote for others is to reward them for producing content you value. If all you do is vote for yourself then you may get a token amount by milking the system, but you will be burning your reputation and not paying those who produce stuff you like.
Finally we have something that discourages milking behaviour and more on producing good content. +100!
If I have 2 accounts and the one account vote the other?
Thank you for elaborating on the new algorithm. I believe that this succinctly states what is the purpose (quality content creation and discussion) and how this new algorithm will reward it.
I am confused, however, regarding how the reward for voting other people's comments work. Is it still the same mechanism where you contribute some amount of rshares based on your steem power available at the time? Is it still a maximum of 5% of your total available voting power?
How is it determined that a user receives activity rewards in voting? I see that you say that there is no difference if you vote more than once in 24 hours versus only voting once. What does that mean with respect to your payout if you vote for one post that is more popular than another? You get rewarded the same regardless, from what I understand?
In essence, it is then more beneficial to comment on the posts you like with meaningful discussion.
Just trying to wrap my head around the new rules.
Basically, I want to know how your activity reward is determined if you say, vote for Post A and then Post B and give the same amount of rshares to both, but that Post B ends up being more popular than Post A.
For simplicity, say that you are the first voter -- vote order doesn't matter so much anymore, but for the purposes of understanding what's happening, just assume you are the first voter -- for both Post A and Post B. Then say that at the end of the day, Post A has 100 worth of rshares and Post B has 200 worth of rshares, with you contributing, as an example, 10 to each.
Does the order matter here as to which post is used to determine your active reward? Is the maximum chosen in this case, i.e. you receive activity reward based on the reward given to the post that receives the maximum number of rshares?
Say on the second day you vote for Post C, and Post C only has 50 rshares at the end of the day.
What happens if on the second day Post A receives an additional 200 rshares, do you then receive the additional activity reward from Post A or Post C and the corresponding rewards based off of your weight?
Or, are any of these scenarios invalid because I do not understand the new reward scheme ?
I think the active reward will be calculated like this: when reward for a post is paying out, fetch a list of accounts which has activity in last 24 hours (no matter if the activity is on the post), distribute 15% of total pay among them -- weighted by the accounts' current SP. If it's the case, there might be lots of calculation and data-update, perhaps performance would be an issue.
Activity reward is not tied to any particular post.
Oh, that sounds better.
Assume every account has a
next_activity_reward_timefield, it gets updated tocurrent_time + 24 hourswhen a new activity occurs AND only if the account's next_activity_reward_time is in the past; and maintain a globaltotal_active_spvariant and a globaltotal_pending_payoutvariant; on every new block, pay the accounts which passed the payout time. It seems the result would be a little different than paying every 24 hours though, perhaps there are better implementation.The change is a bit drastic, although the eventual outcome is intended as an improvement, we need to see how this change will effect rewards on the 4th, maybe @arhag can enlighten us. To me the 4th is just another day when it comes to steemit, but it seems we started the fireworks a bit early for some. This might be our biggest debate yet, just like the one that happened a little while ago with bitcoin's block size.
" only accounts with simple posting authorities consisting of a single key can qualify for activity rewards. There is still the n^2 rule which means that voting for yourself will cause you to earn below-average returns compared to voting for others." These statements by @dan need some clarification. or anyone who understands them.
I'll be honest, I haven't looked very deeply in to the way rewards are given right now, but I love hearing that the crew is looking to make things as fair across the board as possible. Hopefully the way this breaks down will lead people to posting posting bigger, legitimate, thought provoking posts. Right now it seems like some things are just quickly typed up and submitted just to make a quick buck. I'd like to see the focus on community involvement like you've described here.
I really like the adjustment, especially because of the ability for everyone to get higher rewards and also to encourage the discussions. Great work.
Awesome guys!
It's like you've taken all the best comments and then wrapped them up with your philosophies.
For me this comment sums it up:
I definitely felt that and I stopped voting on comments that I really wanted to because they were being kind about my writing or making a great point.
Thanks guys, I really think this is a fair way to go, obviously there will still be some that won't be happy; you can't please everyone, all the time. But this is the best solution I think, activity is the one for me.
I have just done an article with 16 responses, which is the most I've had and that pleased me more than higher earning ones with no comments.
Quick Question: Does my accidental down vote of my own post matter less now; did it ever matter?
Sorry 2 QQs: Does cancelling a down vote still reset rewards and is that resetting just for the day; or all of time up till that point?
Cheers
CG
Changing a vote only affects (affected?) the rewards for that post only. Changing a downvote does not matter at all, except for the voting power.
Thanks Pharesim; do you know how it affects voting power?
5% like any other vote, but that is due to change soon