Steem Follower Checker - new scoring for the follower network strength

in #curationlast month

The last time I updated the scoring method in the Steem Follower Checker was back in June. Since then, my goal has been to add some friction for older accounts, because many followers go idle as time passes.

This weekend, I took my first step towards correcting for that inactivity by implementing an adjustment for the followed account's age. ChatGPT graphed the adjusted curve like this:

The actual age is on the X axis (horizontal) and the adjusted age is on the Y axis (vertical).

Basically, the idea is that every four years, the weight of a single day doubles for calculating the average number of new followers per month. I also experimented with two years and three years, but four seemed to yield the most reasonable results.

Here's what the theoretical scoring distribution looks like, new and old:

New curve, October, 2024

Old curve, June, 2024

Not too much different, which was actually the goal. The main changes that I wanted to see were just in the long-lived accounts that have had time to accumulate large numbers of inactive followers.

Here's a heat map comparison of the new vs. old scoring distribution based on real-world posting activity (one day in June, two days in October).

New scoring method (based on adjusted new followers per month and median follower reputation)

Old scoring method (based on new followers per month and median follower reputation)

And here is a frequency bubble chart comparison, also based on actual posting activity (one day in June, two days in October)

New scoring method (based on adjusted new followers per month and median follower reputation)

Old scoring method (based on new followers per month and median follower reputation)

Finally, here is a side by side comparison of the score groupings. The June method is on the left, and the October method is on the right. Again, this is based on actual posting activities.

Each column is colored according to the median number of followers per month in that particular "bucket".

In short, it now appears to be easier to avoid a score in the lowest range, easier to get a mid-range score and harder to get a top score. The overall median and average edged up slightly, but not a lot. Unsurprisingly, the adjusted number of followers per month dropped from about 43 to 37.

Looking forward

Ideally, of course, this score would involve checking the recent activity level of each follower, but that doesn't seem to be computationally practical, so heuristics are needed.

I can already see two improvements that are needed with the current method:

  1. It doesn't seem to penalize old inactive accounts with large follower networks enough.
  2. It seems to penalize old active accounts a bit too much.

So, for my next update I intend to combine it with the method that was previously suggested by @moecki to focus more on new followers within a certain time window through use of SDS. I don't want to lose sight of all followers, as some may still be active if an inactive account returns to the blockchain. Instead, I intend to augment the score with a 2nd factor that incorporates recent activity.

Conclusion

Overall, the idea of this scoring is to reward authors who combine two behaviors: (1) grow their follower network; and (2) help their followers to succeed.

I have already implemented the new method in my autovoter, and I'll keep my eye on it for a few weeks. Assuming no problems become obvious, I'll merge it into the master branch of the Steem Follower Checker at the end of the month - October 31, or thereabout.


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.


image.png

Pixabay license, source

Reminder


Visit the /promoted page and #burnsteem25 to support the inflation-fighters who are helping to enable decentralized regulation of Steem token supply growth.

Sort:  

Upvoted. Thank You for sending some of your rewards to @null. It will make Steem stronger.