Hacker News new | past | comments | ask | show | jobs | submit login
Tell HN: The Submissions System is Broken
136 points by tomerico on Oct 19, 2010 | hide | past | web | favorite | 49 comments
6 days ago I submitted the following story:


It's a blog entry of an iPhone developer, sharing an interesting story on his marketing attempts of his app. He shared actual sale number, amount of effort, and a tactic he used which actually got him featured on the app store. This story got even more interesting after I submitted it, because he became so successful to hit #1 best seller in the US app store for a moment, and now he is #5.

When I submitted the story, it didn't get noticed, and quickly went past the first page of new submission without a single up vote. This is of course natural. As there are many stories being submitted.

The problem is - you will notice it has 6 up votes now - this means that 5 additional people has tried to submit this story to share with everybody in the past 6 days. But the system completely ignore this fact, and the story doesn't even have a chance to hit the front page again - just because it missed the chance once...

There are many possible solutions, I can offer at least 2:

1 When a story that didn't hit the front page get resubmitted, it hit the new submission page again, as it was just submitted.

2 A small part of the front page will be dedicated for the 5 latest submission, this way new stories will have thousands of views, and will have more chance to not get ignored.

An idea to address #2: http://imgur.com/HocGF.jpg

The HN Toolkit userscript [1], amongst other things, shows a preview of the new submissions page to the right of the usual front page links.

[1]: http://userscripts.org/scripts/show/25039

Someone posted on HN a couple weeks back about the story ranking algorithm. PG mentioned that there were some hidden tweaks, but the basic idea was a function of number of votes and age of the article since submission (a "gravitational" term of sorts which drops stories over time).

This type of algorithm explicitly penalizes stories with a pattern like that which you mention. If a post gets little interest initially but later becomes very interesting, it can't possibly do well in the ranking algorithm.

It would be more realistic if the ranking algorithm took into account the age of the votes more than the age of the article. Recent votes can pop an old article back to the front page, but not if that article had a ton of old votes as the "center of mass" (average age) of the votes would still be pretty old.

Here's a fairly recent post, and in the top comment pg exposes those hidden tweaks: http://news.ycombinator.com/item?id=1781013

i think the reddit system of featuring one new story on the front page would be good. Show a story to a 100 users...if it doesn't get an upvote...retire it, if it does show it to another 100...if a story doesn't get 5 upvotes...retire it.

Yes I agree. I've found that the reddit system has a lot better chance of bubbling up interesting articles irrespective of who submitted it.

Unfortunately, from my experience, HN has become a bit of an echo chamber in the sense that certain posters have a lot better chance of having their articles noticed than the majority; irrespective of the relative value of the individual articles.

I don't have data to support; but with the current system it feels like HN is increasingly pushing users into two extremes - new users (who have an extremely difficult time of getting anything noticed) and longterm posters (those who whatever they post gets noticed and promoted). Long-term this is not conducive to a healthy, growing community.

"new users (who have an extremely difficult time of getting anything noticed) and longterm posters (those who whatever they post gets noticed and promoted). Long-term this is not conducive to a healthy, growing community."

^ Digg

"an echo chamber in the sense that certain posters have a lot better chance of having their articles noticed"

It's also an echo chamber in terms of ideas. Why is it that when a woman does a reverse job application she gets 4 upvotes, but when a guy copies her two years later he's 'a genius' and gets 450+ upvotes?


Just throwing this out there...

    Why is it that when a woman does a reverse job
    application she gets 4 upvotes, but when a guy 
    copies her two years later he's 'a genius' and 
    gets 450+ upvotes?
Because it was two years ago, and the link is to a landing page.

Your implication that sexism is the cause is pathetic.

To clarify I don't think it's sexism per se, but rather the fact that she's non technical and not part of the tech community. I think if she had been a Ruby programmer and the design and content of the site had fit the norms of the HN community then it probably would have done just as well. (Adjusted for vote inflation.)

1. This will just lead to more spam. Submit, re-submit...

2. Won't work. They are recognized as new stories and they change quickly. Nobody wants to waste his time with "possibly" uninteresting stories.

The solution is to prohibit people with less than a karma threshold, are new, didn't contribute a lot in the discussions from submitting stories. Also, limit other members to 1 story per day, this will give equal chances to all HN users.

That rule change is not about stopping someone like tomerico from submitting stories (he has a lot of karma and submits only sporadically), it’s about reducing the volume of stories on the “New” page and giving them more exposure, right?

I was already about to write a retort because at first I didn’t understand the mechanism by which this rule change would improve HN. Reducing the volume on the “New” page seems like a good idea, though.

If there is 500 eyeballs per hour for the new page and 30 new submissions, each one get on average 25-30 visits (a user hits a link or two).

If we reduce new submissions 5 or 6 times, each story will get 150-200 hits and thus more coverage, enough to get it to the front page if it worths the FP.

I miss the info on how many eye balls the article got. If a submitted article gets only one upvote, is it because it didn't have readers or because people didn't find it worth the upvote ?

I recently submitted an article [1], it got 4 upvotes and translated to about 200 views.

[1] My Startup Ideas - http://www.puremango.co.uk/2010/10/ten-ideas/

Here is yet another way to tackle this.

If an article is voted for first the first time after time X of not being voted for, it goes back on the new page as of the time of that vote. I don't know the aging algorithm that is used on the main page, but make it be one where the age of the article doesn't matter, but the age of the votes for it do.

This would also handle the case where someone dives through the new submissions page and votes for an article that was posted several hours ago.

This is compounded by the speed by which new articles scroll by in the "New" view. Right now the oldest New article on page #1 is 58 minutes. This means that I can't honestly upvote more than, say, 2 articles out of that list (since I wouldn't have time to read them).

It seems like the mechanism that puts new articles in front of people is at least part of the problem. Since they go by so quickly these days, maybe different sets of readers could get different "samples" of the New list, allowing those items to be around longer (but seen by fewer people). Since the HN readership is growing, I think it might be fair to rely on the statistical sampling this would provide.

Option 1 would result in a lot of reposts getting back to the front page and option 2 would be bothersome or redundant for users that frequent the new submissions page.

I don't know how the current algorithm works but there has to be some exchange rate between independent submissions over a relatively short period of time (multiple days) to upvotes over a shorter period of time (multiple hours). I assume this is already part of the process and that your specific example falls outside of what the algorithm considers to be significant.

EDIT: Clarification

Wrt option 2, the new submissions page already contains stories that are currently on the front page, so it's already a bit redundant. Doesn't bother me that much.

A couple of hours ago I made a comment here about the importance of giving a good article a good title so it does get read in here. In comparison at this moment the original post went from 6 to 25 points (thanks to this post), this post and it's 'better headline' had 90 points, the post I resubmitted has 104 and is second on the home page right now.


The new post completely won over thanks to a better title. Food for though.

My initial response to this was that maybe it could get resubmitted with the boosted number of upvotes, so it would have a slight edge over new stories being submitted at that time. In addition, you could have this only effect stories with less than 10 upvotes, to prevent people from griefing the system. (spam submitting the same link over and over/necroposting old content with lots of upvotes by reposting it and giving it just a +1 for your resubmit).

Hate that I missed that as I think Trainyard is a great app...

I have a few observations about how I use HN and maybe there are others using it the same way.

-- First, the "Tell HN:" or "Ask HN:" is a real attention getter for me. The first post didn't have it and, because of that, it was easy for me to skip it in the submissions

-- Second, I rarely visit the home page or even the "New" page and instead I opt for the "Classic" page (the page featuring submissions that have been upvoted by a person who has been on here for 365+ days

-- Third, your quote that the post 6 votes because "5 additional people has tried to submit this story" isn't exactly accurate. It has 7 votes now - I just upvoted it.

-- Lastly, what time did you submit it? There is clearly a bias towards certain times here.

"... The Submissions System is Broken ..."

Yes it's broken but HN has always been broken to some degree ~ http://www.flickr.com/photos/bootload/398269769

But is the core functionality broken? Not really. These type of problems are edge cases, a PIA, but not killer bugs. For pg, I imagine HN is sort of a positive side effect of developing Arc & YC. Fixing this kind of broken but working feature, the downside.

I think two useful pieces information could be added to the ranking algorithm: the ratio of clicks/view and the ratio of upvotes/clicks - basically the conversion rates of an article, which should be more telling than just the total number of upvotes. If a story is read by 10 people and nine upvoted it, it's a much better story (although with a bad title perhaps) than a story which was read by a thousand people and upvoted 10 times. If a submission gets less upvotes and clicks because it comes in at a time with not too much traffic or comes with a bad title these factors could offset the penalty it receives.

It's not necessarily trivial to count the views of an item, since the ranking quickly changes, but at least the number of clicks should be trivial. One should also weigh the clicks and upvotes according to rank as the rank itself influences the likelihood of a click or upvote.

Also I'd take into account the number of people who submitted a story - a submission is a much stronger vote of quality then a simple upvote.

It probably did not get the attention you thought it deserved because the title is just not interesting. If you think the content is worth the time, take some of your time to give it a better title so people are attracted to the article.

This does nothing for me: "IPhone's Trainyard developer share his story on how he got featured"

But change it to something like: "From 0 to 450k, an App Store Success Story - With Charts and Numbers!"

And I bet the story ends up in the home page. I'm going to drink my own koolaid and submit the story with such a title to see what happens ;)

Edit: I mean, look at what you did here, you posted this message with a title that people noticed and boom, you now have people looking in to the article that wouldn't have already.

This just leads to another problem though: the first submitter can bury a story by submitting a poor headline.

Yep that's correct, but that's how it is now. I can submit a story with any title I want. If you really want to submit a story that has been buried by a lousy title, there are ways to do so (as I did with this story a couple of hours ago)

"It probably did not get the attention you thought it deserved because the title is just not interesting."

True. However, I don't like the trend of voting up stories because they have interesting titles. It leads to stories about iPhone developers losing out to stories about civil liberties in airports (that has nothing at all to do with hacking).

The trend is not to upvote stories based on their title, the trend is to attract more users into the story and the inherent discussion of it. I never upvote a story I haven't read or at least skimmed over and agreed its worth other peoples time.

There's only 30 slots on the front page at any given time, so something has to give. I'd rather miss good stories than have a part of the front page which is consistently junk.

For what it's worth reddit tried the same strategy but they deprecated the feature.

It's basically the same on Reddit. A couple of times I've tried to submit an interesting new story, only to be told that it has already been submitted and I'm welcome to add my vote and comments to some six day old ghost town of an article.

I like number 1: http://news.ycombinator.com/item?id=1724516

I think that a resubmission should have more weight than an upvote.

Then people could game the votes more easily by submitting in groups, rather then voting in groups.

This is already happening. Some bloggers like ryanwaggoner are openly asking for upvotes by submissions at the end of the blog posts. So, unfortunately, this is something that will always remain broken.

I think that an active HN member "asking for the sale" at the end of a post is clever and responsible marketing.

More so than many blogs' gratuitously inflammatory titles or even having widgets to submit every post to Reddit or Digg or every other site under the sun.

Then you build a system that detects voting rings and other irregularities like too many votes from a certain geographic location without that geographic location being mentioned in the title (protects Hacker News Nights)

Then there would just be a bookmarklet to upvote by resubmitting in which case anyone who cared about their upvote would click "resubmit" instead.

I don't think people care about their upvotes that much.

Maybe there is not enough people checking the new submisson page. This is logic because it takes time and competes with work to do. Though there is an incentive to check the new submissions because chances for a comment in them to be upvoted are much higher because it gets more eye balls. This is not true when there are already 40 comments like in this thread.

I agree. I think HN is becoming a victim of it's own success, and it's UI wasn't built to scale. (aside - that's a challenge to all you folks interested in ux ;) ).

I would say that we definitely need a way of detecting duplicate submissions. And I would also add another suggestion: categories (or tags) for filtering.

I don't really support 1, it's just too easy to abuse.

edit: the ideas about a threshold (even just two or three for now) of resubmissions to put a sub 5-vote submission back on the new list sounds like a good idea that's significantly less likely to be abused.

If you really believe your story is FP-worthy and it doesn't make it the first time around, I suggest you delete the original post and resubmit it a few hours later.

I wouldn't recommend that, it can be seen as spamming. Not to mention that everybody thinks his story is FP-worthy.

Spamming would be spamming. Reposting once is not spamming.

I think people have different incentives. If you spent time hacking together something and want to share it, I think you'd be more likely to think your story FP-worthy. If you're just submitting something that you read, much less so.

Or resubmit from deeper fathoms of the omnivorous echoplex.

Although... HN never rewards reposters, I mean never. If it has already been submitted from a different site it won't get a single upvote, ever. Well maybe not ever, but extremely rarely, but that's because the community is not very fragmented.

I'd disagree with that. In terms of who is browsing the site now, versus 4 hours from now, it is very much fragmented. If you make the frontpage, it's likely you'll be viewed by a large enough portion of the community to be punished for reposting. If you don't reach the frontpage, I think it's OK to delete and try again.

I meant if you post the same article from a different source.

Lets not forget that time of day of submission can lead to a getting buried or at least getting less exposure than it would otherwise...

well heres sumthing new.. I cannot submit to HN via my mobile. Any answers? Tx.

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