Live data from Hacker News

Hacker News Analysis

blog.echen.me

21–30 of 62 posts

Re: Hacker News Analysis

#21
post #7

You're working from a different definition of "good comment" than Graham is. The concern with "bad" comments isn't simply that they exist, but that they have deceptively positive scores.

...which is a problem with a crystal clear cause: if you allow purely democratic voting, then you'll end up with really popular crap reaching the top of the pile fairly often. Imagine how useless Pagerank would be as an algorithm if it considered links from every drooler with a domain name to count for the same amount as those coming from the frontpage of Yahoo - that's essentially what you're doing when you tally up votes on a site The only reason it's not worse than it currently is is that there's no real incentive for people to game the system, so we don't end up with massive comment spammers. But we still have douchebags, fools, and all other sorts of sheeple, and in large enough numbers they end up diluting the votes that should count the most.

Several solutions are probably obvious to everyone here and I won't go into them, but the point is, you'd have to be willing to throw fairness aside in order to fix the problem - if a brand new account's vote is worth the same amount as mine, which is worth the same as tptacek's, then something is wrong.

Re: Hacker News Analysis

#22
post #7

You're working from a different definition of "good comment" than Graham is. The concern with "bad" comments isn't simply that they exist, but that they have deceptively positive scores.

...which is a problem with a crystal clear cause: if you allow purely democratic voting, then you'll end up with really popular crap reaching the top of the pile fairly often. Imagine how useless Pagerank would be as an algorithm if it considered links from every drooler with a domain name to count for the same amount as those coming from the frontpage of Yahoo - that's essentially what you're doing when you tally up…

Your last suggestion seems to make things less democratic understandably, but how would it be implemented?

Re: Hacker News Analysis

#23
post #15
post #2

Interesting analysis. I strongly disagree that what he defines as "Quality of Posts" measures quality of posts, though.

If coming up with an objective measure of quality was feasible, the voting system could be scrapped entirely, and instead of making a site that allows users to submit links, pg could have written a web crawler.

But web crawler doesn't have an important input: how HN users would vote on a particular submission.

Re: Hacker News Analysis

#24
An interesting analysis (which only pg could do) is to multiply each vote by ln(user's karma) or something similar. That could arguably deal with "noobs who don't follow the rules" problem.

The above analysis will assume, though, that the problem of decreasing quality of votes (or discussions) is caused by newcomers and not by old high karma users getting nastier.

Re: Hacker News Analysis

#25
post #22

Earlier quoted context omitted.

...which is a problem with a crystal clear cause: if you allow purely democratic voting, then you'll end up with really popular crap reaching the top of the pile fairly often. Imagine how useless Pagerank would be as an algorithm if it considered links from every drooler with a domain name to count for the same amount as those coming from the frontpage of Yahoo - that's essentially what you're doing when you tally up…

Your last suggestion seems to make things less democratic understandably, but how would it be implemented?

On a hunch: Run Pagerank (the original thing, i.e. just the stable distribution of the random walk) on the graph of who-upvotes-whom. Then tally upvotes weighted with the upvoters pagerank. It's probably safe to assume that the pagerank of a given user changes rather slowly, so this shouldn't be a too big problem to actually compute (and needs to be update at most every few days). Also, to favour recent history a bit one could decay the wheight of older upvotes in the who-upvotes-whom graph.

Now that upvotes are non-public information, people can't know their own rank (at least not easily), so this looks at least halfway resistant to gaming.

Re: Hacker News Analysis

#26
post #22

Earlier quoted context omitted.

...which is a problem with a crystal clear cause: if you allow purely democratic voting, then you'll end up with really popular crap reaching the top of the pile fairly often. Imagine how useless Pagerank would be as an algorithm if it considered links from every drooler with a domain name to count for the same amount as those coming from the frontpage of Yahoo - that's essentially what you're doing when you tally up…

Your last suggestion seems to make things less democratic understandably, but how would it be implemented?

Your last suggestion seems to make things less democratic understandably, but how would it be implemented?

Weighted votes based on karma. Or restrict the number of votes a user can cast based on karma or account age.

A problem with these kinds of things is determining what does or does not work and picking up on unintended side-effects, and doing so in a timely manner.

For example, if "young" accounts have very limited votes available will new users be less likely to stick around long enough to become more active voters? How long would it take to see this? What if by the time you recognize an undesired side effect you've already sent the site down the road of ruin (or something)?

For example, how long will pg keep comment vote hidden? How many people have found this to be sufficiently detrimental that they have, or soon will, leave HN and not come back?

How do you craft meaningful site experiments like this while keeping risk to a minimum?

Re: Hacker News Analysis

#27
post #19

I've been thinking lately about running some analysis of my own, but couldn't find a data dump and didn't feel like scraping. Why has the dump been pulled? I'm thinking of classifying entries as temporal/atemporal. Temporal are industry news, press releases, rumors and alike. Atemporal are reference materials, essays, howtos, analysis, links to libraries/frameworks/products. Essentially, I feel that while temporal en…

It's not obvious to me how to automatically classify stories as (a)temporal. My understanding of the way submissions work is that if Bob submits something that Alice also recently[1] submitted, then it just gets counted as an upvote for Alice's story.

Then, of course, there are the "this happened recently, which inspired me to think about some broader trend". For example, during the AWS outage, Coding Horror had a story about Netflix's Chaos Monkey[2], which one could easily classify as both temporal and atemporal.

[1] For some definition of "recently" that I haven't yet figured out. [2] http://www.codinghorror.com/blog/2011/04/working-with-the-ch...

Re: Hacker News Analysis

#28
post #22

Earlier quoted context omitted.

...which is a problem with a crystal clear cause: if you allow purely democratic voting, then you'll end up with really popular crap reaching the top of the pile fairly often. Imagine how useless Pagerank would be as an algorithm if it considered links from every drooler with a domain name to count for the same amount as those coming from the frontpage of Yahoo - that's essentially what you're doing when you tally up…

Your last suggestion seems to make things less democratic understandably, but how would it be implemented?

I like the idea of upvotes and downvotes being multiplied by your avg score. tptacek's votes would be worth 8.85 at the moment, bermanoid's 3.5, and mine 2.92. Newbies would come in at 1.

Re: Hacker News Analysis

#29
post #22

Earlier quoted context omitted.

Your last suggestion seems to make things less democratic understandably, but how would it be implemented?

I like the idea of upvotes and downvotes being multiplied by your avg score. tptacek's votes would be worth 8.85 at the moment, bermanoid's 3.5, and mine 2.92. Newbies would come in at 1.

Wouldn't you want newbies at 0.01 (or some other value below 1) until they have a reliable score?

Re: Hacker News Analysis

#30
post #7

You're working from a different definition of "good comment" than Graham is. The concern with "bad" comments isn't simply that they exist, but that they have deceptively positive scores.

...which is a problem with a crystal clear cause: if you allow purely democratic voting, then you'll end up with really popular crap reaching the top of the pile fairly often. Imagine how useless Pagerank would be as an algorithm if it considered links from every drooler with a domain name to count for the same amount as those coming from the frontpage of Yahoo - that's essentially what you're doing when you tally up…

The site can generate a lot of traffic to a submitted URL, so there is certainly a temptation to game getting a URL onto the front page.

I think that an informal oligarchy was established, certainly with respect to downvotes and flagging. The question is why is it breaking down. More troubling are the number of quality commenters who seem to have departed.

One hypothesis I have is that many on the leaderboard (and may of contributors I value who aren't on the leaderboard) don't seem to have any plans to apply to YC, they just want to take part in a community of hacker entrepreneurs. If more effort is not given to establishing a richer community model that is not managed as an adjunct to the accelerator they will probably continue to drift off.

Post reply on HN