zlacker

[return to "How Hacker News ranking really works: scoring, controversy, and penalties"]
1. sethba+Xk[view] [source] 2013-11-26 13:51:31
>>jseip+(OP)
I find it really disheartening to learn that any article with "NSA" in the title is pretty severely penalized by HN's algos. This seems like one of the seminal issues of the decade, for this community in particular.
◧◩
2. danso+Dq[view] [source] 2013-11-26 14:42:22
>>sethba+Xk
I think we have to take the referenced article with a grain of salt. If I read it correctly, there is nary a mention of the flagging function. So, to put it bluntly:

Correlation does not equal causation

NSA articles may be more often penalized...but they are likely more often to be flagged. It's not that people here are defending the NSA or anything, it's that people have often said that they are tired of reading NSA all the time. And so, their reaction may be to flag those stories, giving them a penalty.

The referenced article does not really account for that, AFAIK

◧◩◪
3. kens+as[view] [source] 2013-11-26 14:54:15
>>danso+Dq
The behavior I saw with NSA articles is they were already penalized by the time they hit the front page, while other articles were up for minutes or so before getting penalized. This implies (but of course doesn't prove) the penalty is automatically applied, not the result of flagging. (And I mention flagging three times in the article.) Also, there is code in the published source to automatically penalize articles based on "rallying cry" words in the title. It's possible people are flagging NSA articles right off the "new" page, but it really looks like they are getting automatically penalized.

Edit: Several people said I should have said more in the article about flagging. There's no way to distinguish between penalties applied by administrators and penalties due to flagging. So there's not much I can say.

◧◩◪◨
4. danso+Pt[view] [source] 2013-11-26 15:09:21
>>kens+as
Ah, my bad...I was reading on a tablet so didn't do the Ctrl-F-find-the-word to back up my assertion. I think I'm likely biased into thinking that flagging plays a larger role, and thus expected to see it more in your analysis, whether it's the reality or not. But yes, capturing the state of article ranking as they hit the front page would help control for independent flagging behavior. It's possible that people are flagging on the New section but I have to doubt that...those would be HNers with a bit too much time on their hands.
[go to top]