Ask PG/HN: Is there something changed in the rank/sorting algorithm of HN?
It occurred to me that a couple of articles are hitting up the top places pretty fast with just 1 other upvote (eg 2 points total).
This used to never be the case, is this a little experiment? Is it based on Karma?
I happen to see some spam articles get pretty fast at the top yesterday/today.
(
21 comments
| 21 points
| on HN
)
Combine that with the strong spam filtering and fast moderation, and you have a winning formula. If something lame hits the front page, you can expect it will be [dead]ed quick.
Edit: changed strength of above writing somewhat. My memory is bad and I don't have contemporaneous records, which should make me wary of saying things like "almost always."
Loading any page when I'm logged in has started taking forever. If logged out, it doesn't take long to load.
For example:
This comments page loads in 981ms when I'm logged out. If I log in as axod, it loads in 7.5 seconds.
Maybe it's just my account :/ Hopefully it's not a 'moderation feature'. Maybe due to a reasonably high comment count for my account? idk but it makes HN pretty boring waiting for page loads.
Is anyone else seeing this behavior?
Not here, but over the past few months, the app has been consistently fickle about loading up the userinfo page and the threads page. Sometimes it works. Sometimes not. On really high-load days, it was taking 7-15 seconds and then just stopping to show the page.
Paul said a week or two ago (I think) that he was tweaking the GC because of the load on the box. Not sure if that could be a factor or not.
Formerly, things would tend to load either almost instantaneously or not at all. E.g. comment pages. Now, there is not infrequently a second or three's delay, but then the page does indeed appear. I find this much more preferable.
http://news.ycombinator.com/threads?id=pg
instead of back to this comments thread. Thinking I'd hit a glitch, I backed up and tried again.
So, I did hit a glitch in the redirection, but after my comment was processed and accepted.
My understanding is that the ranking of front-page items is based on [# points]/[time since submission], which is obviously very large immediately after an article is submitted; this wouldn't normally matter much, but if the high points/time value is being cached, a low-point submission could stick around on the front page until the cache entry expires.
On the other hand, one of my recent ASK HN posts (http://news.ycombinator.com/item?id=1167298) got 7 or 8 votes in the first hour but never made it anywhere near the front page, which seemed a bit weird.
Since PG says the code hasn't changed, I'd assume the number of people online and submitting during the given period has a large impact on whether any particular story will hit the front page.
http://news.ycombinator.com/ask
Here's what it takes to define such an operator, in case anyone's curious:
Anyway I'm pretty happy how HN sorting algorithm works, meaning even with two votes you can go to the front page if it's quality article will remain on front page otherwise you are going to the second page.
This is smart system because every article has a chance to go to the front page and on the front page the "human factor" who will decide the destiny of an article. Not like many other sites that you need "million" votes to go to the front page meaning quality stories can be lost because people visiting the front page is much greater than people visiting upcoming articles page.