Hacker News Reader

> I'm pretty sure that's unequivocally false. [...] I could spell out how

What makes you sure? If you know something we don't, we need to hear it—the community relies on us to get this right. It's also personal: I've poured countless hours into combating vote manipulation, and if you know my code isn't working, you bet I want you to spell out how.

The upside of a public discussion outweighs the downside because (1) if you tell us things we don't know, we can start investigating for them—it's what we don't know that's worrisome; and (2) community members might respond with valuable knowledge of their own. That's also why we haven't buried the OP (which, in case users are wondering, we do know about and have lots of data on).

I'll email you.

For what it's worth, I don't think HN is in any immediate danger, nor do I know of any active, large-scale vote manipulation happening currently. But to say something like "vote manipulation on HN does not work" is a very bold claim.

Ok, I look forward to it. Edit: still waiting!

I understand that HN's ranking algorithm isn't 100% disclosed because people post startups here and this forum needs to be a fair environment. I say this to make clear that I do get the security ramifications of revealing too much.

But it sounds like you have a bit of understanding when it comes to this topic, which I'm very interested in understanding more about generically speaking. I've always been curious how vote manipulation detection _actually works_, and the kinds of things that can be done to combat vote hacking of various kinds.

I say this as someone vaguely considering the idea of making a small HN-like forum at some point (still figuring out the why and how).

I'm pretty sure upvotes from different users are unequally weighted when it comes to getting a story onto the front page, and that upvotes from fresh users aren't counted with nearly the same weight as someone who's been here a while and has a track record of predicting which stories will perform well on the front page. There've been times that I've given a story languishing on /newest its 4th upvote and it pops onto the front page, and there've also been times that I've seen stories get 10+ upvotes in short succession and get stuck on page 3. It also wouldn't surprise me if receiving a lot upvotes from newly-created accounts, or accounts that have all upvoted the same things, ends up tripping the voting-ring detector.

If you look at how PG originally created the algorithm in Arc this behavior makes a lot of sense, especially when alongside it is the explanation that HN will cancel out votes that look shady. (I could be wrong, but I believe this is the case).

Basically the algorithm looked like:

Score = (p-1) / (t+2)^g

p = points (the -1 is to negate the point of the original submitter)

t = time

g = gravity (or some gravitational constant)

So the p value matters a lot; especially when there's a smaller t (a new story). This gives newer stories a chance to bubble up with fewer points against the ones that have 1,000 points but have been on the front page for days. It's a pretty ingenious algorithm, actually.

It makes sense, given that formula, that even if all p values are weighted equally, moving p from 3 to 4 would make a big difference, and would be able to push something new to the front page. You'll see them now and then with 2-3 upvotes on the front page, because they're very new and the front page is relatively stale.

Anyway, I'm ~100% sure spammers can get around the vote ring detection, but I don't think that's happening on HN in a major way.

These were cases where the submission was both older and the point value was lower than other submissions that were not on the front page, though.

It's possible that flags, voting ring detector, or spamkiller were negatively impacting the other story, but it definitely isn't a straight votes & time formula anymore.

> I just don't have the time or the money to shell out and buy IPs.

You need a Hacker News account to upvote, and obviously having a large number of new accounts upvote something would trip a voting ring detector.

> Furthermore, if you produce something solid, all you'd really need is the first <10 upvotes to get on the front page and let it catch on "organically."

Much easier said than done. The median score of a HN submission is 1-2 points last time I checked. (the discoverability of HN submissions on /newest is an issue, but that's another topic)

And if it's actually solid, it's not the market for vote brigading.

I agree manipulation does happen, but it doesn't mean it works and instead silently penalizes the submission to death.

> You need a Hacker News account to upvote, and obviously having a large number of new accounts upvote something would trip a voting ring detector.

Correct, but what if I gave myself a couple or few months to prepare accounts? Do a search on blackhatworld (or another spam forum) for "aged accounts" and you'll see that you can generally buy them in tiers - the older they are the more expensive they are.

I'm not saying it's worth it or I think it's something people should do, but saying it's not possible or doesn't work is absurd.

I think it'd be fairly obvious that something suspicious was going on if a large number of dormant "aged accounts" just happened to login for the first time in months to upvote a specific thread.

Except this is why there is an economy of aged accounts....

It directly affects the cost of the promotional service