Tweaking curator rewards algorithm for better content discovery
@anonymint recently made a post asking Would Steem fail if every blog post of a lady putting on her makeup is rewarded $26,000?
In his post, he covered one of the consequences that result in the way curation currently works, namely that there is not enough diversification in content type in successful posts.
But there is another problem that result from the current ranking and reward algorithms:
With the way the current algorithms work, the best strategy for curators is to simply follow the most popular posters, and vote on their posts as soon as they appear, hoping to get there before everyone else does. Since these posts are almost guaranteed to become successful, they reap the rewards.
There is very little incentive to find new, unknown posters who may have very good content, but may not become successful because no one knows about them. Why take the risk when you know there will be 100% successful posts? Especially when it takes a lot more effort to find high-quality content from unknown posters.
Curators are not rewarded, but penalized for taking "risks" voting for relatively unknown content creators.
This isn't the way curation should work. The purpose of curation is to allow discovery of great content that would otherwise be left undiscovered. A top poster who consistently makes the top trends do not need to be discovered.
This is why I believe we need to take into consideration of another factor in calculation curation rewards: how established the poster is. If the last 10 posts of the poster all netted over $1000, you can expect that the next post will probably be quite good as well. Thus, there is much less of a benefit to the overall community for curating established posters.
Curators should be rewarded for putting in the work to discover new content. On the other hand, established posters shouldn't be penalized for consistently creating high quality content, either.
I'm not sure if there are more recent changes, but this seems to be the way posting and curation rewards currently work. According to that,
"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"
My proposal is that we create a new variable E, denoting how established the poster is on a scale of 1 - 2, with 1 being completely unknown, and 2 being an extremely established poster (exact definitions to be determined).
The actual values for content + curation rewards would be calculated thus:
0.4 * E = rewards for the author of the post
0.625 * (1 - 0.4 * E ) = rewards for comments
0.375 * (1 - 0.4 * E ) = rewards for activity
In this way, we can give weight to how much value the curation actually provided. Established posters will be able to get a bigger share of the pie, because even without the curation, they would still be near the top. There is also more incentive for curators to discover content from posters who have yet to establish themselves, and for the new posters, they get the benefit of being discovered. It's win-win-win.
Very interesting proposal, I think you might be onto something here. Getting rewarded for finding diamonds in the rough would be a nice incentive for browsing and voting on new posts.
Here is a visualization of the proposal:
And a display of varying E values (labeled x on the table).
Y1 is the author percent, Y2 is the comment percent, Y3 is the activity percent
I think the spreads are nice, but am curious though about how you came up with the 0.4, 0.625, and 0.375 multipliers. Arbitrary, or specific reasoning?
I am also curious how you can come up with an arbitrary metric for "establishing" a poster. Some variables I think that would be worthy of considering include: Account Age, Votes per Post, and Steem Earned per Post.
Great graph and chart! Admittedly the numbers were somewhat arbitrary (I merely took the original ratio between commenting and activity rewards, and did a +/- 0.2 with the original author reward). For me, the key is the point of the post, not any particular set formula. If some of the more important people ends up seeing this post, I'm sure there will be a great discussion about just exactly how to tweak the formula. It'll be much better than anything I can come up with.
In terms of what metrics to use for how "established" a poster is, I think Average votes per post could be important. I'm not sure if there should be a cut-off (i.e posts over a certain amount of time, say 2 months would not count) or not. Again, if we can spread this message and get some important people on board, we'll be able to have a great discussion on the specifics, and come up with something better than any single person would think of.
I think vote weight is more important than average vote per post. Obscurity on Steem comes from not having enough whales voting on your post.
very interesting idea and proposition, I would definitively love this to be implemented.
You make an interesting point. If steemit has to successful it has to have a great collection of strong, diverse, and exclusive content. This can only happen if good contributors are discovered and rewarded for their content. Which in turn is possible only of curation system promotes some content digging.
Fantastic proposal! This is the type of post that should be making $1,000+
I actually think the weighting for activity should be much lower. (I assume activity means voting). Like 0.15 or even lower at 0.005. Otherwise automation is too lucrative, and a recklessness for curation is perpetuated.
Thanks for reading. I think it's hard to know just what is an appropriate % for activity and comments. Maybe as time passes, we'll have a better idea.
but I think what is most important is to have a variable %, depending on how established the poster is. A whale poster does not need any curation, and the community gets no value from curating whales. Thus the activity % should be extremely low, and most of the value go to the poster themselves.
This also discourages automating voting on whales.
But for unestablished posters, proper curation should be rewarded, I think. It will allow more hidden gems to be found.