Hacker News new | past | comments | ask | show | jobs | submit login
on Apr 25, 2012 | hide | past | web | favorite

Things I've heard at various points which may or may not still be true:

- Votes by direct link don't really count. So you could see a high total but it's not the actual value being used for the rank.

- Many flags can penalize a post

- Voting rings are heavily penalized

I'm in favor of all of these things because they prevent the most obvious gaming schemes.

Edit: looking at your history (http://news.ycombinator.com/submitted?id=destraynor) it's clear you only submit your own stuff. No one will argue with people self-submitting stuff here and there, but only submitting your own posts? If the content was worthy, someone would post it and it would get enough votes to rise organically.

Fair enough, thanks John.

For what it's worth - I don't have a voting ring of 40/50 people (that I'm aware of :-) )

I don't think the post would have been flagged.

I can't control people voting by direct link, as another commenter alluded to, there are twitter accounts with tens of thousands of followers that tweet direct links. Maybe they're accidentally triggering this, which would be unfortunate.

Re: me submitting my own posts, it's just a matter of ritual really. I write a post, then publish, tweet, HN, Facebook it all at the same time. I also dislike when others submit my posts with terrible titles like "Marketers, this is a good article" or "Read this post about emails".

Thanks for your time John.

> Votes by direct link don't really count.

what is a 'vote by direct link' and why?

A vote by direct link is, I think, what happens if you post a link to the comments section of a article (i.e. this type of page) and ask for upvotes.

so, if someone comes from twitter @newsyc20, @newsyc50, etc. the votes don't count? looks like a bug not a feature

That's my understanding, but I'm not sure.

The sudden drop is a clear indication of down-votes. I've not looked at your specific case(s) in detail, but from the descriptions here it sounds like some few people here have taken exception to you or your material and decided to flag your posts, possibly without even reading them.

So if some of your upvotes come from what the software thinks is a voting ring, and/or if a few people flag your posts, they will drop to #80 or below very quickly.

Several technical solutions have been suggested in the past, but personally, I think it's a mistake for flags to count so heavily in the ranking. I think flags should not affect the ranking, but that sufficient flags compared with upvotes should result in deletion.

Consider. At the moment if a few people take exception to your posts then they will disappear from the front page and get very few views. Vigilant vigilantes can bury items quickly. Worse, spam gets buried, but not deleted.

If an item truly is spam, let it remain visible so that more people will flag it and thus cause its deletion. Simply burying it runs the risk it will never get deleted.

Thanks for trying to help shed light on this.

If this is the case, it would do HN a great deal of good for this system to be made more transparent. I understand that it might not make sense to publish the exact formula or algoritm which could expose "exploits" to those wishing to game the system, but I think it would be useful to at least show how many flags and down-votes a post has received. It would, for one thing, put an end to this speculation.

If it is the case that these extraordinarily fast demotions (#2 to #88 in one step) come from "normal" flagging and down-voting, I would ask: Why the hate? :-) I can say categorically that people LOVE Des' posts—not one gets less that 50k uniques in a week or so. Including the ones that are ripped off the HN home page. If it's simply because Des posts all of his hard work here, I would then ask: Does HN not want good content? And Des simply submitting it does not ever cause it to hit the home page—it takes a lot of votes to get it there. (Which is more than can be said for YC hiring posts…)

I love(d) HN, and it's totally within the rights of YC to do what they wish with it. But situations like this are doing a lot of long-term damage to the trust one can have in it as a source of quality, relavent content.

Disclaimer: I work for Intercom.

It's pretty clear to those who watch and take notice. I'm not the only one to have written about this several times in the past, a quick search will turn up several discussions about flags and rankings.

With the continuing popularity of HN the audience has become much more diverse, so it may be that the majority of people here quite simply don't agree with you. On top of that it only takes a few active dissenters to cause items to drop, as said earlier.

So you say "categorically that people LOVE Des' posts." Well, clearly not everyone. And with the range of interests, that's only to be expected.

For what it's worth, I agree that things aren't optimal, but it won't change. So live with it, or build something better.

OK. It's sad if "that's that" and that a simple fault will be allowed to prevail and damage an otherwise great resource. If it really is the case that a couple of flags can rip the post of the home page, that's a really dumb system IMO—i.e. that many people like the post and upvote it but two people don't agree and overpower them. Doesn't make any sense to me and I'm still skeptical that this is how it works. Thanks again for helping.

You consistently set off the voting ring detector.

Are you sending it out to your friends and asking them to upvote it? I read a pg comment that mentioned there are mechanisms on HN to deal with "voting rings"... So if you submit something and ask a bunch of people to upvote it each time, I believe it does something. I could be wrong, but trying to offer some ideas.

Besides voting rings, another annoyance I have is people creating throwaway accounts to add positive, but content-free, comments a story.. literally minutes later.

I wish there was a 48h delay between the creation of an account and being able to post your first comment.

I wouldn't expect this to solve all our problems, but I think it would help.

I've read the same pg comment. I'm pretty sure that Des is not doing this. Thanks though.

But isn't repeatedly promoting your own stuff spam? I mean if not, then what is?

Leave it to others to decide if your posts are worthy of promoting on HN.

I think the issue is that these posts are interesting, do gather votes but are kicked off the front page anyway.

+1 Except AFAIK Des only ever wrote 2 posts that were in any way "promotional". And even at that, they're pretty damn classy posts. E.g. this is the other one and it's been one of the most popular posts on the Intercom blog to-date: http://blog.intercom.io/automated-emails-customer-respect/ But isn't any post on a company blog a promotional post in some way? (Maybe not.) Disclaimer: I work at Intercom.

Des has submitted dozens and dozens of his own articles. How could say that he only ever wrote 2 posts that were in any way "promotional"?

Moreover, I think Des himself knows the commercial value of all this promotion:


Blogging has been our most effective weapon thus far. We'll give advertising a go soon, just to compare. But right now it looks like a good article that's related to the product drives readers, which we try to then funnel into the site.

By promotional I think we're talking about "selling a product", so when I write a post like this say: http://blog.intercom.io/retention-cohorts-and-visualisations... there is no "Buy now" proposition, it's just me explaining something that I've invested considerable time into researching.

I think everyone wins when this happens.

Because he is concerned that it isn't being left to the people and instead is being artificially dropped. In general this seems like a valid concern because I've seen a lot of "whoa why did this just die" questions over the years. A little bit more transparency might help. I dunno

> it isn't being left to the people

We do know how many people upvoted. We don't know how many people flagged. Perhaps very many flags are needed to drop a story.

I'm sick of good/interesting posts disappearing of the the frontpage. It's not just a case of them gradually falling, but more like going from the front page to 300+ in the space of minutes. It's a shame. I used to like the HN content.

This happens to me every time I submit my own links as well, I have just stopped bothering

I don't think that people are maliciously pushing your content to lower rankings. IMHO I think the amount of stories being submitted and actively voted on are whats pushing your content down further.

For what it's worth - that link looks really spammy. It reads as if you posted it and then got a bunch of friends to astroturf. Probably got flagged a lot. (the comments on HN, not the article).

How many votes up are your posts getting?

If it's only a handful, you're never going to last on the front page for long -- it's part of the standard HN algorithm (something with time/number of votes and maybe something with spread in vote count over time -- at least that's what it seems to be)

Going to the posted url, it shows 46 points in 2 hours. That should be more than enough to keep it on the front page for a while.

44 points in the space of about 20 minutes should be front page material (going by my experience of reading HN).

But regardless, it should slide down , not literally drop on a refresh from #2 to #87.

A flag or two will do that for you.

You have no idea how many flags are needed.

That's a really interesting claim.

Let's perform a thought experiment, all completely hypothetical, of course.

Suppose someone, call them R, has been on HN for over three years. Suppose also that over those years R has, for various reasons, ended up with more than one HN username.

Suppose that R noticed some interesting behavior in relation to flags and rankings and decided to investigate. Waiting until one of his (or her) submissions hit the front page, R then used the other usernames to flag it and watch how the ranking changed. Having done that, R then unflagged the submission to check that it returned to its previous ranking, providing evidence (although obviously not a guarantee) that there had been no other flags to affect the ranking.

At the end of that R would hve a pretty good idea of how many flags were needed to change the ranking of an item.

Would you agree?

Having worked it out, though, R might decide not to make this information public. Why? Perhaps because R respects PG's decision not to make various details of the rankings system public, and chooses to cooperate.

So now, you claim that I "have no idea how many flags are needed." In the light of the above completely hypothetical thought experiment let me ask:

Are you sure?

Added in edit: In response to the replies to the above thought experiment, let me add that "a flag or two" was not intended to be taken literally. My apologies if you thought I literally meant either exactly one or exactly two. I did not. In hindsight perhaps I should've said "a few." I do not, however, intend to make public exactly what I do know about this issue.

> Now, you claim that I "have no idea how many flags are needed." In the light of the above thought experiment let me ask:

> Are you sure?


I can tell you I'm sure it's more than one flag to take you off the front page, and I'm 90% sure it's more than 2.

In reply, I can tell you I've just taken an item off the Front Page with two flags, and then put it back again afterwards by unflagging.

Added in edit: Cool - I report the results of an actual experiment providing evidence in a discussion, and I get a downvote. Fascinating - I clearly still have much to learn about the people who read and react on Hacker News. I have no idea why this should get a downvote.

Could this be a bug in pg's ranking algorithm? I know the internals are kept secret, and is usually in flux, so maybe a recent change caused this odd behavior?

Registration is open for Startup School 2019. Classes start July 22nd.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact