Analysis of artificial deranking of front page HN submissions | Hacker News Reader

Analysis of artificial deranking of front page HN submissions

news.social-protocols.org

3 pointslaimewhisps7 comments

I understand what these graphs are saying about the things they're measuring, but I'm not sure that I understand how it comes to the conclusion that this story has been artificially deranked? I'm no algo expert, I'm just curious how others are interpreting this.

You can see that this story should be #8 on the front page but it was artificially deranked.

I think I understand. It looks like there's a difference between the raw ranks and the actual rank because of some additional algorithm that sits on top of the "raw" algorithm? It sounds like there's additional filtering happening looking at the comments, etc to try to keep people from engaging in internet fights. Is that what you mean when you call the down ranking artificial?

In addition to the flamewar penalty algo, there’s also the second-chance queue algo, and the unlikelihood that all non-staff users have precisely equal vote and flag effect, and that commenting provides a signal at all. I could probably come up with more but best to draw the line at two minutes of thinking about it :)

That assumed "should be" is bogus. By what standard?

Ironically though, the merged thread on this topic happens to be at exactly #8 on the front page: https://news.ycombinator.com/item?id=50101813. It has been on the frontpage for 10 hours and will probably be there for quite a while longer.

Ironically, the link being analyzed by the post is one part of a dupe group, and so all members of it would be suffering extra downranking by attentive folks like me (I was too slow today, though) flagging dupes and reposts in alignment with the guidelines. But the algorithm algorithm site linked here has no context for that, and the page linked under the unrelated neutral title is for only one post in the group rather than the front page. Which is quite unusual, as there are many better links to submit about this site than to a single standalone article, unless the submission’s intentions are disingenuously about the mod actions for that article rather than about the title as written.

And, after commenting here, I see that the mods have responded to the boringly predictable accusations that resulted from checks notes marking an article as a duplicate submission, and their response applies to the unstated accusations underlying this post as well. (I’m not a mod, in case that’s left unclear by the phrasing above.)

So having discovered a new offshoot of that thread in the submissions queue (I read /new a lot these days and largely skip the homepage), yes, I’ve flagged this post, which the algorithm algorithm will duly report: for title misrepresentation, as it’s not a submission about a site that analyzes HN, it’s a submission of a claim (an argument) being made over a single post, except that no claim is presented beyond the title, and without context this isn’t a meaningful submission to me. Write an actual blog post and post that instead.

Copying the mod’s reply here for missing context:

https://news.ycombinator.com/item?id=50109654

Suppression implies intent, no? There was/is no intent to suppress this story. It got penalized by software for various reasons (mostly unrelated to the topic), as well as flagged by users. We've turned that off now.

If, instead of assuming sinister deeds, you (or anyone) would simply give us a heads-up at hn@ycombinator.com, we could fix it sooner. This eventually happened.

I see charts. I don’t see the analysis

Reply on news.ycombinator.com