Yeah, you have been at hive for a long time, and seen many plots on hive related data, but I bet you have never seen a plot like this! It is amazing how much you can do with data these days, and how much the data tells you about behavior of crowd.
Thanks to 
The trouble of any data query at hive, is the fact that the moment you want to go deep into the lower HP domain, you almost hit a wall. This is because there are too many inactive accounts that were created over the years and are now defunct. In order to do any kind of data analysis, such accounts just slows everything down. I also know how to identify those accounts in terms of my simple indentifier: Investors or Extractors. I will just classify those 'dead' accounts as extractors and move on. However, the reverse question is more important. How do you identify an active account? There are many way to answer that and I am not going to get into that debate too much. Let me tell you how I identified these 4349 accounts (which are active) that is plotted above. I was looking for active authors or at least commenters. Here was my criteria:
In order to get author reward, they must create a post. To get 500 HP author rewards, the account must make many posts, just commenting and getting upvotes in comments will not get you 500 HP author rewards easily. The 10 comment cut-off is a low number, but lot of the farmers don't comments much, and I did want to include them. The 6 month cut-off is arbitraty, but I am not interested in people who left a long time ago. We have pondered on them enough, and it is time to move on.
So that was my logic for the selection. I must make sure, that I explain, that whether that number is 4349 or 8000, or 12,000 matters not to me. I just wanted to have a large enough sample to play with so that people can't talk about sampling biases.
I already wrote a wall of text, and the plot is too far above:) Let me put the plot back down here again.
This is a plot of 4349 'active' hive accounts. They are ploted according to their HP rank along the horizontal axis, from the highest to the left (rank 1) to lowest to the right (rank 4349). The vertical axis is HP ploted in log-scale, so that I can plot 1 and 1 million at the same time. The data clouds are:
Overall, KE Ratio increases with lowering of held HP. This is no surprise as held HP is in the denominator. However, the change is dramatic as soon as held HP drops below 200 HP.
Well that 11K average is skewed because of a few accounts with less than 1 HP :). Let me take them out. These are those 20 'authors' by the way.
If I ignore them, for accounts that hold below 200 HP, the average KE = 169.89!
Did you get it? This is a massive dataset. When people holds more than 200 HP, their 'behavior' in terms of hive etiquette is about 30X better than those that holds less! We have excluded, worst of the worst from this calculation.
Here is the disclaimer: Since KE is a ratio, its effect is really prominent when the held HP gets very low, as it is in the denominator. But that said there is a trend of the exploitation among low HP accounts that can't be ignored.
When I first published the data of top 2500 hive accounts back in August 2024, I included every account. It was described in this post. That first plot in that post is below:
The problem with this previous plot is it included all accounts regardless of if they post on hive or lot. A lot of them were inactive, some were investor account, some were curation only etc. So the plot was very noisy. This time around I am focusing on recently active accounts (6 month prior, we can use a different time frame if we want), who had written posts, created content on hive blockchain, and were rewarded as authors.
The immediate difference you might see, is there is no account below 57 (which is the lowest on this very large sample). Most of the accounts are above 60, and show a nice trend, and large clustering. 50 - 100 K HP and 65 - 80 reputation. Reputation here is a function of upvotes received or author rewards. The outliers of these plots are far and few and highly visible.
People who got a lot of rewards and sold it all: kingscrown, anomadsoul high on the Y-axis, low on the X-axis
People who don't post much, but likely bought HP, bubke with 92K HP and 0.12, low of the Y-axis, high on the X-axis
People who post a lot, and still held decent HP, like oflyghigh, taskmaster4450, high on both X- and Y-axis
These people are all outliers. I have tried to label as many outlier as I can. We can look at each group of people in detail at a later post. This post is already way too long.
However, if I don't mention one thing from the first plot then I probably won't be able to sleep at night!
Please, I want you to think about this. What is some of your peers are doing? Which vote trail or curation program some of your peers are on? How come they are getting 40X more rewards compared to you?
Let me rephrase that!
We are going to look into this more accurately on a log-log plot. Here is the Table for that. It will give you a lot to think about. Also there is the chart below that leads to the table for your consideration.
| Peer Group HP | Low End HP | High End HP | Factor |
|---|---|---|---|
| 1000 | 600 | 50,000 | 83X |
| 3000 | 1500 | 70,000 | 46X |
| 10,000 | 4600 | 120,000 | 26X |
| 30,000 | 12,000 | 300,000 | 25X |
| 100,000 | 35,000 | 400,000 | 11X |
By the way, this is obvious, but still I should mention it. Since for each 'peer group' the HP is similar, and their total rewards vary dramatically, their KE will also vary proportionally.
As KE = Total Rewards/HP.
Why do you think the total rewards are so much different between the peer groups? Many reasons:
There could be other reasons that I might be missing, do let me know and we can investigate. It is all in the data.
Similarly we can plot the KE Ratio variability. The results are similar:
| Peer Group HP | Low End HP | High End HP | Factor |
|---|---|---|---|
| 100 | 5 | 500 | 100X |
| 300 | 1.9 | 105 | 55X |
| 1000 | 0.6 | 81 | 135X |
| 3000 | 0.2 | 17 | 85X |
| 10,000 | 0.15 | 14 | 93X |
| 30,000 | 0.12 | 5 | 41X |
| 100,000 | 0.1 | 2 | 20X |
| 300,000 | 0.1 | 1.5 | 15X |
Hope this helps the curators if they like to use this information.
