Homepage Payout Distribution, Power Law and Project Curie
A few days ago I've noticed that the fat tail of payout distribution has gotten fatter. This has gotten me curious, so I've decided to investigate.
I have started by gathering data - scraping the Steemit Trending page with all posts above $10. This yielded a little over 300 posts.
Payout Distribution
Source: STEEM Whitepaper, page 22.
Interestingly enough, the whitepaper predicts the payouts to follow a power law distribution.
Within the limited top ~300 post sample, I have found an inverse relationship to be true - most posts fall into the lowest payout bracket. This is not an exact power law distribution, but it looks close. This confirms the long-tail hypothesis.
Project Curie
At the beginning of this post I've mentioned that the fat tail has gotten fatter. A month ago, Steemit was more like a lottery. The homepage would have a dozen or so high paying posts, while everybody else would get nothing. Does anyone remember the $31,000 makeup tutorial?
If you decide to scroll trough the trending page today, you will find a very long list of posts that range beteween $10 - $500.
One potential reason for long-tail fattening may be the community effort to find great quality content - including but not limited to the creation of groups such as Project Curie, RobinHood Whale and the smoothteam.
I would have to do more digging to actually prove the causation, but for now, I have decided to just look at the distribution of posts that have been voted on by the Project Curie.
The posts are reverse-sorted by payouts. Red marked posts are those which have been upvoted by the Project Curie group.
It looks like these posts predominantly belong to the long tail. We can see this again in a more succinct view of the distribution (again, Red = Curie).
Links
Don't miss out on the next post. Follow me.
That $31,000 post was also down to an influx of noobs who didn't understand the system. They thought if they upvoted the big paying posts they'd get a slice of it. Of course they got nothing at all, but the quadratic formula made the authors of the top posts get an even greater % of the pie.
Users have learnt their lesson now and are concentrating on voting new posts. And the number of new sign-ups is dying.
However, if we ever get a second influx of noobs, you'll start seeing $31,000 posts again.
I believe it was mainly to do with the near $5 price of Steem, that same post today, with the same people voting on it would get less than $6000
Cg
But the noob vote cannot inflate value in posts. Even 10000 noobs combined don't have any serious power to increase value in a post - let alone propel it to multi-thousand rewards...
There is a quadratic function at play. Read the white paper. It is not based soley on steempower. That's why a post with two people voting with 1000 steempower each, will earn more than a post with just 1 vote from someone with 2000 steempower. So noobs piling in does help massively.
I think the quadratic function is related to the total SP of all the votes that a post gets. So, IIRC, it should be the same for 1x2000SP or 2x1000SP.
Here is the white paper:
https://steem.io/SteemWhitePaper.pdf
On page 16 it says in order to discourage people from only voting for themselves, the reward is proportional to votes squared rather than votes.
So minnow votes really make a difference - that's why if you vote for a big post, you can see the reward jump.
10,000 noobs will closely equal to 80,000 steem power, of course its serious power.. man
80000 SP would probably be around a dollar or less in voting power. Stellabelle has ~150.000SP and it's a bit over a dollar I think.
dang that is so small of a number though thanks. Now i see why many are not powering up even investing close to 100K probably only ads a value of not even 2 to a post. There is something that needs to be changed, to start with.
This place is getting better and better day by day ;)
Check out @projectnewbie if you feel like supporting it, I am sure it will help the tail extend even further. ;)
The long-tail fattening should be a good thing as more writers get rewarded. This is good for emerging writers
The fat long-tail helps because Steemit seems determined to maintain wide diversity of content (crypto and recipes, etc). IMO, I think specialization would work better initially. But where fat long-tails should REALLY be useful is for older, timeless content. Writers must be able to build libraries that generate income for months/years.
it's going to be good for a period, but if the users - new ones - grow and there are few high-power accounts and investors, then rewards may become too diluted. we need more whales too
Great job, I was wondering myself about this reward distribution. It's definitely getting healthier.
Enlightening analysis... and some hope too..
What was that editor you're using in your screen cap? What are you using to make your graphs ?
I do the analysis in Jupyter Notebook.
The graphs are either in matplotlib, seaborn or bokeh.
Everything I do is in Python.
Perfect! Post way to go steemit we are now growing as a community like it should!
I have seen several people suggest changing the exponent from squaring to 1.75 or some other number. How much would that impact the fatness of the tail?
woah! Just wondering do you know if the projects are also having bad effects on some of the middle guys? I was earning for a period a guaranteed 30-40 on my story but after Curie became public it dropped to 25 and now it's zero...
All those votes were all Curie too. :) You can go back and have a look - we were voting for 2 weeks before going public. The drop to $25 is due to the decline in the value of Steem.
that's most likely the case
good job, man! :)
Thank you so much for putting in the hard work and coming up with this great set of charts. It is nice to see the platform's rewards being more widely spread.