HN front page ranked using only votes from early users (2009) | Hacker News Reader

HN front page ranked using only votes from early users (2009)

news.ycombinator.com

211 pointsibraheemdev140 comments

I think a lot of the comments here are incorrectly assuming that this is a new page. The classic page was first announced by pg in 2009 [0]. He created it because:

> I wanted to see if there had been any visible decrease in quality.

[0]: https://news.ycombinator.com/item?id=607271

Ahh that makes more sense. It felt like a bit of an odd cutoff point haha

I feel that this experiment can prove nothing about the quality. It does show that votes from initial users have roughly same distribution as votes from later users, but only because the initial exposure to the front page highly correlates to subsequent votes. So it rather proves that both group of users have the same usage pattern: mainly watching the (non-classic) front page. It would have been more interesting to see the front page that only consists of submissions from initial users.

I think a lot of the comments think this is the algorithm that HN uses for its primary page. It's not obvious the link goes to "/classic"

How is this helpful? If the HN veterans did not like the 2020 content of HN they would simply not use it.

The results are just showing the particular interests of that group of people who still enjoy the site after 12 years.

It's interesting for historical purposes to see if the tastes of the "old timers" are different from the tastes of the "newcomers."

It would be hard to measure this directly simply by comparing the front page from 2008 to today, because different articles will have been posted on different days, so a direct comparison is imposssible.

By simply seeing what old-timers vote on vs what everybody votes on, you can make a better apples-to-apples comparison. Not a perfect one, of course, because many old-timers may have quit long ago if HN had changed too drastically, and similarly their own ideas of what HN should be may have changed.

i think his hypothesis which isnt tested by the classic page, is that if 90% of the old timers left hn, the 10% who remain could be those who vote similarly to the newcomers, making both pages appear similar, even though a significant group of old timers may have voted differently.

I have a 2008 account. I do think HN was a bit better earlier on, but I still enjoy using it.

However, HN is also effectively a monopoly. There's really nowhere else to go to find the same type of discussions.

Not http://lobste.rs/ ? I've seen people suggest that from time to time.

Here is its climate models in Fortran discussion: https://lobste.rs/s/zlwgmm/why_are_climate_models_written

Lobste.rs is restricted to computer related content.

On Hacker News you will find anthropology, politics, philosophy, startup advice.

It's possible that someone who thinks that Hacker News was "a bit better earlier on" favours that, especially as (as pointed out elsewhere in this very discussion) those are noticeably lacking from the page at hand. So let's see what the answer to my question is.

Lobste.rs is a "need to know someone" service, so it has limited membership.

Dicussion on Hacker News is not qualitatively better than on any number of well moderated technical subreddits or niche trade forums, and most of the articles posted here are cross-posted from Reddit and other sites. The only real differentiating factor is that Hacker News aggressively downvotes humor and polices for tone.

Every now and then threads pop up asking about alternative HN-like forums, here are some:

https://news.ycombinator.com/item?id=20023209

https://news.ycombinator.com/item?id=16342192

https://news.ycombinator.com/item?id=23402797

I strongly disagree. No subreddit has dang (and the even less appreciated sctb) as a moderator.

Sadly we don’t have sctb anymore: https://news.ycombinator.com/item?id=21930336

> If the HN veterans did not like the 2020 content of HN they would simply not use it.

I’m not sure how accurate that is. For example, I strongly dislike new Reddit, but I feel forced to go there as the niche communities are still far more populated than their respective forums scattered across the internet.

I’m not sure HN has quite the same holding force, but I’ve yet to see anything that feeds my curiosity in the same way.

If by "new Reddit" you mean the UI, you can use old.reddit.com or go into your account settings and enable the legacy UI, if you're logged in.

To give an example, I'd rather not have votes count from people who are upset about this one particular method of ranking. I might start using this view more!

Which is quite helpful. It’s my default bookmark as I prefer the mix.

// This account is 2010; I had been in the earlier cohort under different account. A decade later I still find the early but still active cohort votes a mix of links I prefer.

Also, how much of the ranking is just determined by factors other than user-votes?

I made a similar piece of software to Hacker News about a year ago and I found what was probably an earlier copy of the source code for hacker news somewhere.

I seem to remember that the score that items are sorted by was a weighted sum of the inverse of the age and the votes. I don't remember if the votes themselves were assigned different values based on some context.

Perhaps ask pg? Who knows, he might tell us :-)

(I'm not a Lisp'er, and I've never been particularly fond of Lisp, but I have to admit I enjoyed reading the Hacker News source. It had a lovely minimalism to it)

I remember the day I joined HN in March 2009. The entire front page was full of Erlang articles. Like, almost every single one. I didn’t know it at the time, but apparently that was part of an intentional effort to scare away newcomers who weren’t “hackery” enough.

I thought to myself, if these people are really this crazy about Erlang, maybe there’s something to this website after all and created an account. I have to say, I was somewhat disappointed when it turned out that it was all a ruse!

The say Steve Jobs died, every single story was about that. Not complaining, the day was treated well this way.

In fact, I learned about it from Hacker News! Someone posted it here as a “self” post a good 20 minutes before news outlets or even Apple reported it.

Same here. I’ll never forget walking into my boss’s office with the news. We ended up grabbing our iPods (remember those) and going for a walk together in our music. That was a sad day but a special moment in my career.

I was eating at a campus canteen in a research institute in East Asia. And I saw it on TV. It was weird. I remember where I was on 9/11 and for that.

Actually: https://news.ycombinator.com/front?day=2011-10-05

Also look at https://news.ycombinator.com/front?day=2011-10-06

and https://news.ycombinator.com/item?id=3078988

WOw pretty good memory you guys! I see you were watching HN that day. I don't think I even knew what HN was then.

I find it ironic, or fitting, that the 30th story on Oct 6 2011 is actually about Flash, a thing that died relatively soon after, in no small part due to Steve!

I found out while sitting in a hotel lobby at Abode's MAX conference in LA. Watching the news ripple through the tech crowd attending was something.

Can you provide a reference to where it was indicated this is a ruse?

https://news.ycombinator.com/item?id=512145

> day I joined HN in March 2009.

Happy Cake Day!

This was originally announced[0] with the title HN Frontpage ranked using only votes from accounts over a year old

So is there a reason to think now that the cutoff date is Feb 13, 2008, or could it be dynamically counting accounts over a year old?

0: https://news.ycombinator.com/item?id=607271

Even for 2008 cutoff, the classic ranking is still remarkably similar to the main ranking. (Which stupefies me.)

So why bother tuning it to be even more similar?

There's nothing surprising to the ranking being similar.

Old accounts don't use classic, they use the normal HN homepage like everybody else. Which means, they see the same top links as everybody else, and upvote them, like everybody else.

In other words, /classic is remarkably useless. It might have been useful when it was created (11 years ago!), if a sizeable part of those old accounts would use /classic, and only /classic.

I use /classic, slightly less dross. It’s gotten gradually less different through time, perhaps something to the point that others may not be using it.

the likely cause of similarity is the "classic" users browse (and vote based on) the regular front page

> Which stupefies me.

You're surprised that the general tech population is remarkably similar to the elite tech population? :-p

I think it's supposed to be about the "startup" users. So you can't use a dynamic cutoff date.

Nowadays, HN is more of a general tech aggregator, but discussions where some users lament the turning away from exclusively startup content still happen every now and then.

(Personally I'm happy with how it went. Startup stuff bores me.)

> So is there a reason to think now that the cutoff date is Feb 13, 2008,

Damn, March 31 2008 account here. Juts over a month too late :)

I just refreshed both the Primary and Classic front pages and did a quick comparison.

Classic had 14 articles that were not on Primary. The first unique article for Classic showed up at #15. All of the unique articles for Classic were at least 11 hours old, with 5 of them being at least 1 day old.

Primary also had 14 articles that were not on Classic (obviously). The first unique article for Primary showed up at #4. All of the unique articles for Primary were at most 12 hours old, with 4 of them being at most 1 hour old.

This page may have originally been created to investigate a difference in "quality" between older and newer HN users, but it actually does a great job of highlighting the opposite to me. The Classic page is mostly the same as Primary, just slower. Older HN users seem to generally like the same things as everyone else here. I'm sure most of the pre-2009 users who don't like modern HN have moved on.

> Older HN users seem to generally like the same things as everyone else here.

Probably most HN users (both older and younger) mainly vote for stuff already visible on the Primary page. A hypothesis from a parallel thread, not mine.

What follows, the users mostly vote as led by the "trend setters", by whom I mean those who frequent /newest subpage.

Okay, I get the idea here, but... my account is younger than that and I'm sort of miffed by the notion that my votes shouldn't count at all. Furthermore, having the cutoff date be in the past means that I could never earn that right regardless of how much I participated after that.

I would be much more well-disposed to a view that only counts votes from accounts which are more than a year old, or some other criterion which reduces young and drive-by accounts, while still allowing people to eventually count as old-timers.

What I want is a news aggregator like HN but where I can control (possibly have it learnt) how to rank based on network effects. Preferably in an interpretable fashion if any machine learning gets involved. Heck, I would even be happy for it to be partially probabilistic to get me to “read across the aile” at times.

I've thought about that, however I worry about it turning into what Facebook does and put you in an echo chamber (with the occasional "across the isle" item thrown in to "make you angry"). But I still like the idea, just not sure what the best implementation would be.

Some ideas I've had include having multiple up/down vote buttons for different criteria -- one that says if you agree or not with the content, and another one that says if it is a well reasoned and interesting post. That way I can pull up things that people in my cohort think is a good post but also disagree with content wise. But this requires that people are reliable able to judge something fairly that goes against their view.

Then I'd like to see "bubbles" that show which cohort(s) I belong to and which other ones are out there so I can choose which to read at any one time.

"Aisle," both of you, like in the senate chambers.

Slashdot moderation has this, with separate voting for insightful, interesting, funny, etc, and the corresponding toggles on the view side if you want to prioritize or exclude particular categories.

It already is an echo chamber. Except that you have zero control over the content you see, so you're stuck in someone else's echo chamber. It can be really fucking frustrating, which is why I closed my main account after 8 years.

you're still here tho, so why close the account?

I don't really want to be here. This account will be gone too as soon as I close my browser. I won't make another one unless something very urgent comes up.

I always thought HN (or reddit, youtube, etc), with a Bayesian filter would be interesting.

I'm sure there's a good amount of obscure content that doesn't get appreciated because it never reaches the people that would enjoy it.

That's certainly the case but how would a Bayesian filter help?

So tiktok but for news?

Smart personalised and decentralised search with comment aggregation and sharing.

It shows you what you're interested in, not what adtech, or social media barons, or VCs, think you should be reading for your own good. Or for "engagement."

Something something echo chamber? Probably. But it's not as if we don't have those already.

You can already do that. Just click hide on posts you don’t want to see.

The issue with having a model that does it for you is that your going to immediately start fighting against it and its decisions.

Hate to be the stereotypical pedantic HN commenter, but fwiw this normally works, I believe there's an exception with Ask HN posts where clicking "hide" will not fully hide it, at least last time I tried.

Wait, have I been on HN since 2011 without realising this? There is nothing in the FAQ at least [1].

[1]: https://news.ycombinator.com/newsfaq.html

I have always avoid pushing “hide” since I assumed a new submission would pop up at the bottom and I never wanted to go beyond the top 30 as I feared I would stay on the site for a bit too long.

Thing about things that happened before you is you can never be a person who experienced them.

That said, could be implemented as a rolling decade. The term is “classic”. For cars that takes 25 years. Ten for tech seems fine.

That would be accounts from 2011 and before.

Still curious, what had happened to HN on Feb 13th, 2008?

I just scrolled to its frontpage[0] and the top story was “Immediate social bonding between strangers is highly dependent on mimicry”[1].

Any other explanations why `https://news.ycombinator.com/classic` turned back to this date?

[0] https://news.ycombinator.com/front?day=2008-02-13

[1] https://news.ycombinator.com/item?id=115094

On google and wikipedia, HN's launch date is listed as "February 19, 2007". I presume this is to screen for that first ~year of accounts.

Funny, since I remember lurking for years before making mine!

I upvoted this comment, but it doesn't count.

Your vote counts in the "primary" ;). You are part of selecting stories for the main page. Overwhelmingly, "classic" voters will vote for things that are already visible on the regular main page.

It probably started as a fun gimmick rather than an actual feature people are supposed to use?

Why do you care about a random view on a website?

Your vote counts on the main homepage.

> a view that only counts votes from accounts which are more than a year old, or some other criterion which reduces young and drive-by accounts

Are you sure you know what the motiviation for accounts older than 2008 is (are you sure it's reducing young/drive-by accounts?)

Have you thought about what that motivation might be?

I assume that it's to filter for people who have been around a long time. If there's some other specific motive, I'd love to hear what it is.

It was an experimental feature slapped together by pg in 2009, and has been largely forgotten about ever since. The motive then was to see if there was any difference if you filtered out newer users. There wasn't then, just as there isn't now.

Its continued existence is just a historical artefact.

As far as I can see it's not linked from anywhere in the UI, and is only occasionally referenced in comments.

Nothing to be miffed about!

That's the what, not the why

Also, nothing has been removed, only added. You can see more of the data which was hidden before.

Why was the arbitrary date of Feb 13, 2008 chosen?

I think a mechanism with filter-by-date might be interesting but that date of 13 years ago is arbitrarily excluding most 30-something year olds (all but the precocious 18~low 20s that created a HN account?)

It isn't arbitrary, it is one year after HN was created.

Are there enough active and voting accounts to rate anything adequately?

It's like calculating market quotes only from deals by old enough participants.

Thanks - I had no idea. My question is less about the particular date (an anniversary of the plaform) and more about the year, though

20008 determines more than 'classic', then, it determines 'early adopter'

edit - I see your other comment about /classic being from 2009. That makes a lot of sense!

Meta enough this post was #1 for me on HN, making it the biggest drop of between regular and Classic HN.

At first glance, on the /classic page,

There is less:

- pop-culture & humanities

- notifications about momentary outages across the web

There is more:

- hard-science topics

- algorithms

- posts about VCs

But overall, the ranking is much the same, and all the differences you see might simply be because there is only a handful of votes for some pages from old-timer accounts (ie. the quantization noise is pretty big).

> quantization noise is pretty big

Seems to be true. I have an account which is apparently from 2007, so included for the classic page. When I upvoted the article on place 29 and refreshed it jumped to 28. When I upvoted the article that was now in place 29 it jumped to place 24. Votes on the normal homepage don't seem to have anywhere near that effect.

That the ranking is the same seems a bit of an early assumption to make, without seeing all or more of the data.

In addition, on the /classic page,

There is less:

   - Up-to-date tech

There is more:

   - Nostalgia/Parenting content ( A Calvin and Hobbes 
     guide to parenting? )

It’s enough different to prefer/bookmark/use it exclusively if that’s your bag.

- Rust

So, some form of elitism?

Any specific explanation for ignoring 12 years of newer accounts? To me that’s potentially a whole generation of new engineers’ comments/questions/insights that are being ignored, again for no discernable reason.

If there was a spam problem (or similar) I could understand, but this just seems arbitrary.

Again I would just want to understand the reasoning for it, as it could be legitimate.

/classic is 12 years old too - pg set it up to see how the front page differed based on 1 year old accounts and all accounts.

It was a quick experiment in 2009, and soon forgotten by most. It's not linked from anywhere in the site's UI. It doesn't seem intended to be used anymore, and seems to exist as a historical artefact. Certainly nothing for people to be offended by.

Was there any discussion about that experiment at the time?

Again, mostly curious as to the reasoning, as this kind of thing would tend to go against fostering a community, which is what the idea behind HN seems to be.

There was this discussion:

https://news.ycombinator.com/item?id=607271

Nobody seemed to have a problem with it, and quite clearly the community has grown a lot since then.

Agreed, it should count votes from high-karma new individuals.

What's remarkable to me is that the ranking is almost the same.

If you used

https://en.wikipedia.org/wiki/Bootstrapping_(statistics)

to sample possible top pages you would see #1 and #2 are pretty stable but around #10 or so you see some variation.

So it's not clear to me that the oldsters have different preferences from the youngsters.

When you aggregate preferences it converges to a statistical average of 'voters'; typical users like you and me are already an aggregate of preferences (there is music that i like, software i like, business i like, etc. that are basically disjoint) like the body thetan aggregates that Scientologists say we are.

If you look at the top submitters on HN that are exceptional in the volume of articles they post you see their preferences are very typical and it is hard to tell a difference between the % of articles that they submit on (say) "deep learning" and that appear on HN as a whole.

It's like the way that almost any stock market portfolio you make is going to converge on the SP&500 if you add more stocks and wait long enough.

More salient, however, is the distribution of articles submitted in 2021 as opposed to 2008; "deep learning" was talked about hardly at all in 2008 even though Geoff Hinton was giving talks at the time where he was sketching the idea out.

It is not obvious to me that oldsters have significantly-different preferences than youngsters (w.r.t. chance) and I'd put the burden of proof on those who say it exists.

There's an assumption that older accounts somehow have better or different taste, which I find incredibly questionable.

I think it's more likely that they just happened across the site sooner, and happened to stick around.

I'd like to see the frontpage but sorted by votes from accts with a lot of comments, or a lot of karma. That would be more revealing

It is supremely elitist and very marketing-y (use an impression (older users were better) over using real-world data).

dang's comment on this 3 years ago:

https://news.ycombinator.com/item?id=16438990

Amusingly, it seems to re-surface every now and then.

Yeah, it does seem to pop up every now and then :) Many of the newer users don't know it exists until someone brings it up again.

Should be renamed /gerontocracy

Okay, how OP discovered this? Is there a some kind of index/glossary of HN features where we can look for?

Can we do this by other criteria? I.e. another year or points? The opposite of the classic could be also interesting.

> Is there a some kind of index/glossary of HN features where we can look for?

Yes there is an un-official list of undocumented hacker news features: https://github.com/minimaxir/hacker-news-undocumented

That's great, thank you.

It would be interesting to see how many votes are cast by accounts from that range, rather than just sorting it that way and displaying votes from all users.

I thought I was pretty early to Hackernews (August 23, 2008). I'm curious to know what I missed that happened in Feb 12/13 of 2008!

Turns out "classic" had 10 of my clicked links - while the normal "new" had 5. Will use classic from now on.

Same difference for me, but most of the clicked links are from one day ago. It mostly just seems that the classic page has a slower turnover (due to less votes).

Missed me by three days.

How many of those people are still around?

Apparently a lot, if you compare the vote counts on the "normal" homepage with the "classic" homepage. Actually the difference is so small that I highly doubt the "2008" claim in the title - I think it's still counting accounts that are over one year old!

Obviously nobody who was an adult in 2008 is still able to operate a computer or comprehend the latest tech trends.

This is clearly aimed at tapping the wisdom of the savvy investors who have made a bet on vintage hn accounts as the latest digital store of wealth, indicator of wealth and conferrer of legitimacy. If you were smart then you spent the last decade picking up these heirlooms in estate sales and recovering old logins from fossilised magnetic tape.

Nice idea. Could be better if this threshold is implemented into the user profile and let the user decide.

I really hope someone gets it into their head that they're going to analyze this and perhaps visualize the results in some lovely way. I would love to know if the difference in voting patterns can be characterised and we can learn something from it.

I understand the desire to look at the input for the original users, but I think in the spirit of fun and the tradition of Star Trek TNG there should also be a view that only starts to count after the start of the first year - it could be /beard.

You could expand the cutoff date by modeling lookalike audiences. A 2018 account that votes on similar things to a 2008 account might be admitted. This affords the moderator an easy “exploration” spigot they can tune up or down.

It's neat and number one for me as well; an obvious signaling

Prior to only a few years ago, I considered all website comments to be throwaway and never bothered to sign up or comment.

It's sortof a new exercise to share my thoughts with the group.

It'd be very cool if you could set the classic date in your profile.

Right — and it could perhaps default itself to the date you joined, since that's a moment when it seems likely you liked what was on offer.

That does introduce the risk that people actually use this. Now, this page is a gimmick, a historical artifact. If it were actually used by more than a handful of users, that would be right alienating for new users, especially if it's a fixed date that you have to manually update.

Since this seems to be the week for me to express unpopular opinions, as a recent (2019) user of HN, I welcome this, and have bookmarked it.

Why? Because it seems to offer a much better signal to noise ratio.

The Frontpage rank I'd like to see is a PageRank. Quality is a transitive property; I want to see what is upvoted by those who are upvoted (comments as well as posts)!

It's nice to know what the greybeards find interesting. As an almost greybeard it would be interesting to see what the young'uns are upvoting

I'd be much more interested in seeing such a ranking for comments in somewhat controversial articles.

I just checked - created my account 5 days before pg did the "classic" experiment.

I'm the problem - not!

Can we get one the other way round?

That's like only people above the age 70 elect the president of the United States.

suggestion: instead, just rerank using weighting by age (and other factors?)

Dang, even my account isn't that old. October '09.

Welcome to HN!

Doesn't seem to be working. I see many submissions by accounts created after that date.

Is there some reason why a person downvoted this? When I visit the link, I see links submitted by accounts created in 2019.

>ranked using only votes from early users

The ranking, not the submissions, are from the old timers.

I would love to know why downvotes happen as well.

Whoops. I could swear the title was changed.

Thanks for pointing that out!

That's one hell of an old cut-off

what is the significance of that date?

This sounds entirely reasonable. A bunch of web sites went the route of having a bunch of smart, reasonable people get together and make something real.

When those grew, they became popular with idiots, with spamming/SEO/astroturf, and/or taken over by extremists in some direction (at which point, everyone else heads for the door). The web sites became cesspools of crap.

I'm fine being excluded if that means HN continue to have high-quality content. If the pre-2008 accounts start falling off, it wouldn't be too hard to find a pool of high-quality contributors from 12 years of comments, upvotes by pre-2008'ers, etc.

Indeed, I wouldn't mind having explicit tiers, even if I am in a lower one.

Would love to see this for comments as well. I swear some years ago, I used to be able to find some really interesting stuff browsing the "comments" link now I'm more likely to find throwaway_xyz posting flamebait takes.

I don't think it'll work for comments. But maybe, maybe their upvotes can count more than one vote. Or their downvotes (i think it will create more readable comments)

Yep - it's an interesting twist at least. If HN continues to grow, we all know where it ends. And there's a reason we're here and not there.

Filtering the comments tab on karma would be good.

Roughly the same as your suggestion of course, since if an early user is still commenting, they almost certainly have accrued a lot over the years, but it would also allow newer entrants that have had well enough received (or enough well received) comments to qualify.

Karma doesn't work the same way.

Karma tends to:

- Promote extremism and groupthink. Try posting any comment which isn't far-left on reddit (outside of the far-right forums, which will be the opposite). You'll be -50.

- Be gameable for SEO and spamming.

- Target the mean. Try posting anything particularly erudite on a reddit or a late slashdot.

I kinda like the crowd which set up HN. It's diverse enough, intelligent enough, and works for me.

Not particularly useful because almost all the old timers are still looking at the same normal homepage that everybody else is.

Counting only votes from /new would be interesting.

so, old people?

Old people?

Let's say that the youngest you could be (without being some sort of child prodigy) interested in this content is ~8, that would make the youngest person with an account that old maybe 21 years old.

That sounds super old to me :-)

The only interesting difference is less posts about Facebook get upvoted.

this is peak HN shtick, straight from the heart of startups "oldest members are $BETTER because they invested early in the idea" or some other ideological BS like that...

Reply on news.ycombinator.com