Before I get into the details of the metric, itself, let me cover the reasoning behind it. The reasoning goes something like this:
The goal of the community of Steem curators is (or should be) to correctly rank all Steem posts in terms of value.
In this context, "value" does not mean "payout". It's an abstract and subjective ranking. In ranking the posts, the goal of curation is to get the "payout" near the "value".
At a high level, the value of a Steem post is (or should be) primarily determined by two factors:
The quality of the post.
This includes aspects like originality, writing quality, surprise factor, relevance, information content, attractiveness of the presentation, etc., and it is best measured by a human, an AI, and/or some combination of the two.
The reach, virality, or engagement of the post.
Ideally, we could measure this with view counters, but that information is not available on the blockchain (also, view counters can be gamed), so we need a way to estimate it. @realrobinhood estimates it by counting resteems and comments.
The "Follower Network Strength" is another way to estimate this in a way that lets the curator attempt to predict virality as a function of the author's historical ability to build a follower network.
The benefit of "Follower Network Strength" in contrast to measuring comments and resteems is that it (hopefully) gives an earlier estimate of a post's potential for virality.
Note that I am not including factors such as token burning or club status in the post value. In my opinion, these should be viewed as levers for audience building, but they do not directly impact the value of a post.
As blockchain usage patterns change and possible improvements are discovered, the "Follower Network Strength" metric will need repeated adjustments.
Accordingly, I have set a goal to publish a new version at the end of June, and then quarterly after that.
If curators adopt this metric and it starts to affect authors' payouts, changes will become controversial. So, transparency is required.
The current "Follower Network Strength" score is based on two values that can be easily pulled from the blockchain:
The number of followers
The median reputation of followers
These values are combined into a score that ranges from 0.01 to 1.414 (square root (2)).
The two main improvements that I think are needed for the end of June are:
Follower networks are comprised of active and inactive followers, so all follower counts are not the same. For example, someone who raised 2,000 followers in six months should probably get a higher score than someone who raised 2,000 followers in six years, since a higher percentage of the 2nd author's followers are likely to be inactive.
It is too easy for an author to get a maximum score.
I have known about the problem of inactive followers for a while. However, the naïve solution is to check the follower accounts' recent activity, and adjust the follower counts accordingly. Unfortunately, this involves numerous network calls, so I believe that it's unsuitable for a browser plugin. So, I was stuck for a while without any feasible ideas.
Last week, I had another thought, though.
What we can do without adding much network activity is check the age of the author's account, and then average the number of followers into a time-based average over the life of the account (i.e. new followers per day, new followers per week, new followers per month, etc.). Then, we can replace the follower count with the time-averaged follower count in the initial calculation.
If we do this, then the author who collects 2,000 followers in six months gets a higher score than the author who collects 2,000 followers in six years. Additionally, if an account goes inactive and their follower count stops growing, then their "Follower Network Strength" will gradually decline.
Making it more challenging for an author to receive a top-score
The naïve solution is just to play with the minimum and maximum settings and look at how they change the heatmap. I had actually started down this path before I thought of the time-averaged follower count. However, implementing the time-averaged follower count means that all parameter settings will need to be changed. Therefore, these changes would become obsolete.
Instead, what I'll need to do is to start almost from scratch with the min and max parameters after replacing the total follower count with the time-averaged follower count and then tune the values according to the updated heat maps.
Summary
In conclusion, the planned changes for June are as follows:
Change "follower count" to "time-averaged follower count" as one factor of the "Follower Network Strength" calculation.
Tune parameters to award top scores to a smaller proportion of accounts.
After soliciting feedback from ChatGPT and Google Gemini, one point that I also want to emphasize is that this metric is not intended to be a standalone value. Instead, the idea is for curators to incorporate this metric as just one factor of many that feeds into their voting and following decisions.
Thank you for your attention, please let me know if you have any feedback or suggestions.
Thank you for your time and attention.
As a general rule, I up-vote comments that demonstrate "proof of reading".
Steve Palmer is an IT professional with three decades of professional experience in data communications and information systems. He holds a bachelor's degree in mathematics, a master's degree in computer science, and a master's degree in information systems and technology management. He has been awarded 3 US patents.
Visit the /promoted page and #burnsteem25 to support the inflation-fighters who are helping to enable decentralized regulation of Steem token supply growth.