How Well Does Your Number of Up-Votes Explain Your Author Payment?
When you upload a post, at what volume does the number of your up-votes begin to have a significant statistical association with the size of your author payment? The association tends to become strong primarily when the number of up-votes reaches well above 100. This is a finding from an examination that I made of 263 articles that appeared in the “Trending” menu late on September 30.
By studying these posts, I was able to group them into 53 classes, and the association just cited was tested within the classes that have a large number of articles. These classes are listed at the end of this post.
The two charts that follow feature two of the classes. The first is called “Social cohesion and social support in the Steemit community”, where there were 73 posts. I named the second-class “Help for Steemit authors”, and it contained 33 posts. Here is the first chart.
The numbers that you see in the horizontal axis are simply identification numbers for the articles. On the left vertical axis is the scale for number of up-votes, and on the right side is the scale for payments/rewards. I sorted the observations according to the size of the vote.
In the first chart, you see that the curve for the sorted number of votes (colour red) rises gently up to about 75 votes, and then it ‘screams’ upward. Basically, up to 75 up-votes the payments are ‘all over the map’. The good news, however, is that they have a base value close to 50 Steem Dollars (I guess).
Keep in mind, however, that these are Trending articles only. One of the interesting questions to raise is what was the total of non-trending articles over the time frame covered by the 73 items of the first chart. And if we collected a decent sample of those non-trending articles, where would the distribution of payments be concentrated?
There is another important caveat. Although it is instructive to isolate these two variables for examination, while looking at the articles one by one as we go from left to right, we should be holding constant other factors that help to explain the level of the author payment. If there is substantial interest in this post, I would try to do just this, which would make analysis much more sophisticated.
Coming back to the Trending articles in the first chart, we see that as the volume of up-votes starts to march upward very smartly near the right side of the curve, the central tendency of the payments also moves up quickly; but we continue to have a very wide range between the top and bottom values. Thus, we clearly need a more sophisticated analysis, where other explanatory variables are taken into account.
On the whole though, the chart is suggesting the message that if you are lucky enough to get your article into the Trending collection, you might get some big dollars (say $100-plus) at quite low volumes for the number of up-votes; but a good prediction that you will receive big dollars should include the assumption that your volume of up-votes is in the ‘100-plus area’, and you will be ‘in the sweet spot’ when your volume of up-votes approaches and passes the two-hundred level.
Let’s move to the second chart, where the posts fell into the class called “Help to Steemit authors”. Here we have only 33 posts.
The two curves follow each other more closely than is the case in the first chart. Generally, the rising trend of the sorted number of up-votes is associated with an ‘embedded’ upward sloping trend curve for the payments. Do note, however, that the wild swings in payments that we saw in the first chart seemed to be re-appearing as we move above 100 up-votes in the second chart.
The general message of the second chart seems to be broadly similar to that of the first, except at the curve for payments seems to be more sensitive to modest volume changes in the case of the second chart.
As an author, the key bottom lines for me are the following. First: when I am getting less than 20 up-votes, I am going to need some ‘special forces’ to get big dollars (i.e. in the $100-plus area). This, I'm afraid, is bad news for authors like me who tend to write on rather specialized subjects that have naturally small audiences (sigh)!
Second: I am going to need much more than the average number of up-votes for my type of article. This means that I should first establish what is the community of interest for my article, and then find out the typical distribution of up-votes for articles aimed at this community. Even if my article is lucky to find itself selected as a “Trending” post, I would need a lot more than the average number of up-votes for trending articles in my category, in order for the author payment to be substantial (at least relative to the time investment that I need to make in order to complete and upload the post).
Here is the set of 53 article classes that I derived by studying the said 263 Trending posts. I read far enough into each article to reach a reasonably considered judgment as to what its classification should be, near the end of the exercise when I was becoming quite tired (at which point I made the classification based on the content of the title).
(1) STEEMIT software improvement and extensions
(2) fictional story
(3) analysis of proposed STEEMIT protocol change
(4) promoting a STEEMIT-related event
(5) unclassified
(6) explanation or analysis of a computerized protocol
(7) privacy and security on the Internet
(8) social cohesion and the social support in the STEEMIT community
(9) sport events
(10) sport activities
(11) fashion
(12) lifestyle, personal aspirations and coping
(13) help for STEEMIT authors
(14) applications built on STEEM
(15) promoting non-financial aspects of personal well-being
(16) reflections on successes achieved and challenges met in life
(17) multilingual presentations
(18) volunteer work and giving for and to the needy
(19) promoting Internet-based personal services
(20) reporting external news and commentary about STEEMIT
(21) SMTs: tutorial, technical expositions, and promotion
(22) STEEMIT technical support
(23) guides to STEEMIT features and related resources
(24) promoting financial aspects of personal well-being
(25) STEEM's hardware infrastructure and associated software
(26) music
(27) cryptocurrency news
(28) Art: drawing, painting
(29) photography
(30) ethnic and race relations issue
(31) flowers and plants
(32) Steemit sub-communities
(33) travel and tourism
(34) poetry
(35) STEEM blockchain
(36) new members introducing themselves
(37) governments and crypto currency issues
(38) books
(39) gaming and gambling
(40) true adventure story
(41) curator function on STEEMIT
(42) statistical analysis of trading data and prediction models
(43) coin trading strategy and tactics
(44) Steem Power statistics
(45) contests
(46) Drama and film
(47) recipes and other food preparaton info.
(48) Algorthmic trading, robots, bots
(49) commentray on policices and politics
(50) Getting mainstream adoption for cryptocurrencies
(51) diet and nutrition
(52) managing personal finances
(53) cryptocurency exchanges.
With such a small sample the results are still very interesting. Having content creators like yourself taking the time to research and publish these results is great. I believe that the results would be similar over the whole platform. Would be Interesting to measure reputation with content. Alot of undervalued content being missed and cant get the payout it deserves.
Thanks for your encouraging comment, kevinge. I've tentatively decided to do a series of short articles where I try to bring in other explanatory variables to sort of model the payout level, and continue reporting on results for specific sub-communities of interest.
Also, as I treat each sub-community I will probably take the time get a sample size of at least 50 from the Trending plus Hot selections, and then a much smaller sample from those that did not get selected to be either Trending or Hot.
It will be a long series in terms of elapsed time, God willing, as there is more urgent stuff on my plate just now. However, when the series is fleshed out, say a few months from now, we will have a more strongly evidence-based view on the distribution of payouts and what are its main driving forces.
I do expect, however, to confirm/support some hypotheses already in the ton of writings that already exist on this topic; so my value added will be putting a base of wide-ranging observations behind all the theory or one-person-based thinking that is behind much of the existing literature.
Hopefully, the key results will be in place before new SMT projects try to develop their reward systems. If they merely replicate the social-media focus of Steemit the current system will be sort of okay; but if they try and get more focus on identifying, promoting and rewarding valuable 'green' content in specialized areas, the current system will need major surgery, I think.
Cheers!
Interesting data. It seems very difficult to draw any correlations. I wonder what kind of stats we have on self-voted vs non-self voted posts. Hmmm. I may need to do some analytics. :) Hopefully, my activity helps you increase your audience.
You have been featured on Trump3t, the no-self voter promoter's Update! Created Edition!
https://steemit.com/selfvote/@trump3t/trump3t-the-non-self-voter-promoter-daily-update-created-edition
If you like what I am doing, please upvote, resteem follow. Cheers! 🎺
I just up-voted and Followed you. Thanks for your support!
Thanks and I will follow up your link when I awake in the morning.
For now, I want to say that the article says twice that a more sophisticated analysis is needed, specifically to take into account several relevant variables that we see discussed in many posts.
The key issue, I think, re. the article's usefulness is whether a much more sophisticated analysis would overthrow its main message -- to get good payments consistently you have to tend to have much greater than the average number of up-votes for your particular kind of article.
I wish to emphasize that this is not a hard-and-fast rule like 2+2=4. It is a statement of a statistical tendency, to which we will find exceptions in various cases.
Cheers!
@mrainp420 has voted on behalf of @minnowpond. If you would like to recieve upvotes from minnowponds team on all your posts, simply FOLLOW @minnowpond.
This post recieved an upvote from minnowpond. If you would like to recieve upvotes from minnowponds team on all your posts, simply FOLLOW @minnowpond.
@royrodgers has voted on behalf of @minnowpond. If you would like to recieve upvotes from minnowponds team on all your posts, simply FOLLOW @minnowpond.
This post has received a 1.81 % upvote from @buildawhale thanks to: @lestatisticien. Send at least 0.50 SBD to @buildawhale with a post link in the memo field for a portion of the next vote.
To support our curation initiative, please vote on my owner, @themarkymark, as a Steem Witness
i had sent before 6 hours ago Transfer 1.358 SBD to buildawhale https://steemit.com/poetry/@red-rose/why-me-7-divorce
and until now after 6 hours i did not get my rewards ----WHY
WHY
PLEASE RETURN MY FUNDS BACK
SteemIt Doesnt love you. Stop spamming. Downvoted for being a stupid MOFO who is spamming other peoples posts!
I'm sure you will be refunded, but you are leaving a comment/reply that has nothing to do with the original post. If you are looking for a reply from @buildawhale or whoever. Maybe next time write a post and mention that content creator or reply to THAT creator most recent post. Leaving comments or replies like this is encouraging discussion that has no relevance to the post. Thank you
Congratulations @lestatisticien! You have completed some achievement on Steemit and have been rewarded with new badge(s) :
Click on any badge to view your own Board of Honor on SteemitBoard.
For more information about SteemitBoard, click here
If you no longer want to receive notifications, reply to this comment with the word
STOPAn amazing research to isolate into the form of a variable that links between the initial vote with the vote to the next.
Really, the job you describe gives motivation to berstemit. Thanks for sharing @trump3t
Thanks marzuke. Pl. see my reply to keviginge above for an outline of the roadmap to more and better posts. Your suggests of things I will need to take into account are welcome.
Actually, I expect i will not be able to get data on all of the variables cited as being relevant in the literature; but I hope to get data for enough variables to make the association of curves a lot more interesting.
great post @lestatisticien and very informative for those who were previously unaware of how it all works!!
Tried to give your question a brief answer in our latest episode of Crypto Nights (specifically about how Crypto Nights will encourge SMT development) but we will look to answer it in more broad terms in future episodes.
Hope you can check it out.
Alex
Thanks a lot Alex. I would welcome your suggestions re. things to take into account as the research develops further (pl. see my two replies of tonight above).
I'll check Crypto Nights once more in the next day or so. By focusing on what Crypto Nights can do to help the SMT project to be successful, you will be providing a valuable service. For one thing, there is far too much 'hoopla and celebration' at this point when even the software engineering for SMTs is not not yet completed (according to a new post I saw a dar or so ago).
I'm not saying we should be pessimistic. I am saying we should 'put our heads down and get to work' because of all the loose ends that to be tied up. This SMT project is far more than putting out some cute new software engineering.