This is a chance for moderators to conduct a social experiment: let people speculate in this thread for two days before revealing what it is. Just to see what crazy theories people come up with! :)
Does HN even track how many people view the link? As far as I can tell, there's no JS on the home page other than the not-even-AJAX powering the upvote and downvote buttons, and they're not using <a ping>, so I'm not sure how that information would get back to the servers.
You could simulate it by recording its position on the front page for each impression. Assume that higher positions on the page are seen more. Then your metric is something like upvotes per position-discounted impression.
On the other hand, I'm not sure that dang would get much value out of this kind of metric unless he was trying to tune the ranking algorithm or something like that.
Any normal posts turn red at 51% and above. Ask HN posts never turn red. Went back through the top 500 posts, all of the reds are 51% or above. Might be some other variable that triggers a rare sub 51% red, but I can't find an example right now.
The colors are strange too. Some percents in red, some in gray, but it doesn't seem to be directly correlated with the quantity displayed or the post age.
I really admire you guys for keeping it simple. I have seen too many outages due to piss poor "high availability" design, which would have simply been avoided by not trying to be highly available! Damn, we really need a catchy term/phrase for that (besides ironic)
I've criticised a number of projects for overcompressing the problem domain. Larry Wall's written on this eloquently.
And I'm not criticising HN's server infrastructure. Earlier discussion of style/presentation somewhat. Though for that I'm as inclined to criticise overdesign as well -- mostly on other sites, though arguably table-based layouts are a kludge.
I think it may have been sentiment analysis. The low ranked threads appeared to have a more negative tone, at least during my quick anecdotal survey, and this thread was high ranked perhaps because people hadn't had time to create really long complaints.
It can't be sentiment analysis because some items didn't have any comments but still had a percentage. Some other clues that we have is that the percentage didn't correlate with the color and that pretty much all the recently submitted links didn't have percentages.
In this case, you'd do it based on the upvoting and downvoting ratios of comments.
50 comments, with 15 grey comments = bad sentiment, lots of negativity, non-productive conversations etc. Then come up with a scoring algorithm, with red being volatile or bad sentiment so the moderators can spot them instantly visually.
I remembered there being special options and information for mods from a post a while ago and was able to dig up this screenshot which contains a % score: http://imgur.com/ge3ZBgt
Bit off-topic: At my last workplace they used to push our code for several ecommerce sites to production on fridays, every week, without tests. It somehow worked out.
75 comments
[ 2.5 ms ] story [ 124 ms ] threadNot exactly a foolproof test though. I think you're probably still right.
Perhaps part of their security minded hacker detection system. ;)
Of course, now the percentages are gone.
On the other hand, I'm not sure that dang would get much value out of this kind of metric unless he was trying to tune the ranking algorithm or something like that.
>Posting in epic thread.
Looks like the percentages appeared, then HN went down briefly (server restart, sorry @dang), and now the percentages are gone.
We prefer to call that "restarting the server" :)
I've criticised a number of projects for overcompressing the problem domain. Larry Wall's written on this eloquently.
And I'm not criticising HN's server infrastructure. Earlier discussion of style/presentation somewhat. Though for that I'm as inclined to criticise overdesign as well -- mostly on other sites, though arguably table-based layouts are a kludge.
50 comments, with 15 grey comments = bad sentiment, lots of negativity, non-productive conversations etc. Then come up with a scoring algorithm, with red being volatile or bad sentiment so the moderators can spot them instantly visually.
Some of the percentages appear related to upvoter count.
There is one with 9 upvoters and a percentage of 33.
My guess is it's a measure of clique voting. X% of people voted on another story together.
Not sure, that may be more complicated than needed, but it seems to be a percent measure of the upvoters.
So maybe percentage is how many people viewed the front page but didn't upvote it (since you can't downvote it)...
... A-a-and it's gone now...
Normally I'd say sorry, but this turned out to be so amusing that I guess I'm not sorry. Carry on!
Edit: you guys blow my mind. One of these years we should do this on purpose.
The mystery continues! :)
EDIT: And apparently that % score was discussed here: https://news.ycombinator.com/item?id=7972234