Civitai Writers Leaderboard Changes: Five Suggestions
A question I could not answer
It started with a simple question about the Writers leaderboard. Why does a writer with 111 articles sit below one with 15, whose single old guide carries 59% of their views?
Nothing is wrong with either of them, the guide earned its views. But I wanted to know what board is actually measuring, so I counted. Every public article on civitai.red's API on 22 October 2026: 24,060 articles from 5,633 authors. Then I wrote five boards that score writing a different way, and ran them as a simulation on the same data.
A note before the numbers. To be fair, first idea and first formulas looked well. Then I ran the simulation and came out first in a couple of the groups. I did not aim for that, but I took it as a sign that the formula was rewarding volume. I post a lot of articles, and posting a lot is not the same as writing well. The real drivers of this community are the people whose articles get others writing their own: a challenge or a prompt that pulls in many participants and many new articles. So the boards were reworked. What you read below is the reworked version, not the first one I made.
I still do well on some of the reworked boards in the simulation, which is why every weight is printed below: check them.
What today's board measures
Today's Writers board takes everything an author ever got (views, reactions, bookmarks), puts each kind in one pile, and takes the square root of the pile. It does not care if the pile came from one old article or from many new ones.
So the board mostly measures the past. 52 of its top 100 published fewer than three articles in the last year. That is not their fault. It is just what an all-time pile does.
Who actually writes here
The writing community is real. In the last 12 months, leaving out people who only post announcements:
1,836 people published an article
533 published three or more
101 published monthly or more, and 15 about weekly (52 or more a year)
93 published in at least 6 of the 12 months, and 360 in the last 30 days
Now the whole site, announcements included. Over the last six full months it is about 993 articles a month from about 486 authors, and 147 of them are first-timers each month. Counting everything, the 184 authors with 12 or more articles a year wrote half of all articles. Authors with 52 or more a year get the fewest views per article, 283, when authors who post less often get 401 to 738.
Stories went from under 30 a month to over 140 in February 2025. Over the last 12 months they are about 22% of all articles, and October 2026 alone reached 25%. Announcements and news are about 40%.
Writers are not one crowd. I grouped everybody with 3 or more articles a year by what they mostly publish:
Storytellers: 159 authors, 2,127 articles a year. Most reactions per article, a new piece every 6.5 days, almost no bookmarks.
Guide writers: 202 authors, 1,561 articles. Most views per article, and the only group that gets bookmarked.
Essayists: 75 authors, 614 articles. Musings, mostly a few a year.
Mixed: 65 authors, 1,275 articles. Few people, some of the most prolific.
Release-note writers: 254 authors, 2,896 articles. Model creators announcing their own releases.
Announcers: 104 authors, 618 articles. Announcements without models, and the fewest views per article.
Five ideas for a fairer board
1. Score each article on its own
A pile cannot tell one old hit from a body of work. So score every article by itself, put each count under a square root so the next view is worth a little less, and count only your best articles. Posting more should not win by itself, and the people who post most already get the fewest views per article.
2. Only other people count
The comment counter on an article includes the author's own replies. On one story, 11 comments came from 2 outside people. Counted as people, that story has 2 participants, not 11. So the boards count people instead: readers and participants, distinct people other than the author who reacted or commented, once per article. The author's own comments, replies and reactions never count. Answer your readers because it is good manners, not because it is points.
3. New work counts more
Today's pile never shrinks, and that is where the 52 come from. So on four of the five boards an article loses half its points every half-life: 90 days for active writing and stories, 180 for community and 365 for guides, because a good guide stays useful much longer than a chapter stays new. An old article still counts, just less.
4. Different writing is read differently
People keep a guide and come back to it. A chapter they react to, then wait for next one. One formula for both will always pick a favourite. So each main kind gets its own board: for guides, bookmarks weigh most, and for stories it is comments from different people, plus loyal readers who come back chapter after chapter. Newcomers get Rising Writers, a board of their own.
5. Count what gets others writing
This one comes from watching the community, not from me. What holds writers here together is invisible to every board. On 8 October one challenge article was answered by about 35 writers with articles of their own, within two days, and comments came from 19 different people.
So the fifth board, Community Builders, counts answers. An answer is a later article by someone else that links to yours, and each person counts once. Answers weigh most, writers for whom it was their first article count extra, and comments from different people count too. Getting somebody to write their first article is maybe the best thing an article does.
Five boards, in plain words
Every weight, as each board would print it:
Active Writers. Last 365 days. Half-life 90 days, best 30 count, release notes a quarter.
√(views) / 2 + √(readers) × 7 + √(bookmarks) × 10 + √(participants) × 5Storytellers. Stories from the last 365 days. Half-life 90 days, best 20 chapters count. Loyal readers reacted to 3 or more of your chapters in the last 90 days.
√(readers) × 7 + √(participants) × 10 + √(bookmarks) × 5 + √(views) / 4, plus√(loyal readers) × 15once per writer, then× (0.75 + 0.25 × weeks with a new chapter in the last 12 / 12)Guide Writers. Guides of any age, past year's numbers only. Half-life 365 days, best 20 count, and one guide from the last 365 days.
√(views) / 2 + √(bookmarks) × 12 + √(participants) × 4Rising Writers. First article in the last 90 days. No half-life, best 10 count, at least 25 readers, release notes a quarter.
√(readers) × 10 + √(views) / 2Community Builders. Last 365 days. Half-life 180 days, best 10 count, at least one answer, release notes a quarter.
√(participants) × 6 + √(writers who answered with an article) × 20 + √(of them, first-time writers) × 10
All of this is a simulation on public data. Participants and answers are real, from 3,457 comment threads and 3,365 article bodies from the last 90 days. Reactions still stand in for distinct readers, and all-time counts stand in for the past-year window.
What stays the same
The all-time Writers board stays as it is. Nobody loses what they earned, and the five boards sit next to it, not on top of it. Release notes still count, at a quarter of the points.
I did my best and posted it as a pull request to the site's code repository, the way I see it. I hope someone from dev team takes a look.
What do you think?
Do the six groups match how you see yourself? Should a challenge entry and a reply article count the same, as answers?
What should a Writers leaderboard reward: the pile you built, the people who read you, or the people you got writing?