At the time of this comment this submission has 5 upvotes in its first 22 minutes and it is #5 on the front page.
This means if you can convince just 5 of your friends to quickly upvote your submissions, they will all get promoted to the top of the front page. You can do the same for their submissions. That's a six-person voting ring.
Ideally pg would like to detect these rings and assign your friends' votes a lower weight on your submissions.
I think user engagement with /new needs to be substantially improved to reduce the effect of this. I'm a heavy HN user but only remember to check out /new once or twice a day.
It only takes a handful of people to get a story into a front page position where the votes can rocket upwards This means that the more people checking out /new and voting fairly (on average), the larger these voting rings would need to be to have any effect. (A war of numbers, sure, but "bad" stories would get flagged a lot more quickly too.)
A second problem is that the seemingly low engagement for /new can mean that reasonable items submitted at "bad" times don't make it past 2 or 3 votes whereas similar content submitted at busy times does well. In a good, fair system, that shouldn't be happening, but it's a reason I time my submissions to HN to strike when, I'm assuming, /new is at its busiest (4pm-10pm Pacific seems to work best for me). Without growing the international audience of HN, though, this might be tough to crack!
I think user engagement with /new needs to be substantially improved to reduce the effect of this. I'm a heavy HN user but only remember to check out /new once or twice a day.
Hmm. I probably go to /new 10 times more than I go to the front page. I suspect the problem with /new is that there is so much of it (especially at those peak PST times) that stories don't have time to garner enough eyeballs before they are pushed off the page.
PS. Wish I could un-downvote - I just clicked the wrong arrow due to clumsiness, sorry... Upvoted another one of your agreeable comments to compensate :-)
Not so much. Trust metrics (underpinning Google's PageRank) are pretty efficient at this sort of thing. If people A-Q all vote each other up, but nobody votes them up, their weighted effect is pretty close to nil.
I think the bigger problem is developing an algorithm that reliably solves this problem (and doesn't introduce collateral damage), rather than the computational complexity of the algorithm.
