Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
Is the YC News Karma formula this simple?
11 points by nurall on May 4, 2007 | hide | past | favorite | 24 comments


KPn = Karma for nth submission

KCm = Karma for mth comment

Karma = [SUM(KP1,KP2,..KPn) - n] + ROUND[0.5x(SUM(KC1,KC2...KCm) - m)]


That's what I figured it was, except I always thought of it in these terms:

You get one point every time someone votes up your submission. You get half a point every time someone votes up your comment. You lose half a point every time someone votes down your comment.


How do you actually vote down someone's comment or post? I don't see a downward-pointing triangle, so I thought a only a super-user could do it. Am I missing something?


I believe you have to have a certain amount of karma. Not much, though, since I can see it with less than 20 karma.


There also seem to be other limitations, such as not being able to vote down replies to your comments or stories.


Direct replies or any descendant?


Direct replies to any comments or stories can't be downvoted.


Ah, yes, that makes sense. Thanks!


It should be the other way around, since users rarely create the content of their submissions.


Would make me happy, I comment more than I submit. ;-)

OTOH, actual submissions are often more detailed and well-thought-out than simple comments. So maybe this is the right way after all.


Or it's something new. It is a news site, after all.

I'm a commenter too. I figure once I learn enough and have something significant to contribute, I'll start a blog.


Am I the only person left who thinks that blogs are generally a big waste of time? At least if some product I'm designing fails, I will have advanced my skills by creating it. Is having written a failed blog supposed to make me a great writer?


It will make you a better writer, yes - for exactly the same reasons. The more experience you have the better you become.


Do karma point ever go negative? If they dont (which is my view) then the sum of comments is rounded to 0 if less than 0.

Karma = [SUM(KP1,KP2,..KPn) - n] + ROUND0[0.5x((KC1,KC2...KCm) - m)]

I changed ROUND to ROUND0 where ROUND0 always returns 0 or more. Is there a math way of representing that, that I cant remember?


This should be a philosophical question for Paul. Can the Karma be negative for a user? The trade-off then would be between absolute web democracy Vs keeping the users 'happy' (capitalistic?)


It's still absolute democracy there are just rules that have to be obeyed :-). Anyway the system is heavily (and rightly) skewed to encourage and award participation, hence no down-votes on submissions.


I always get one karma point per comment point above 1. What's the basis for the 0.5 factor?


I've noticed this as well. I start at 1 and anything higher adds one. The karma formula just seems to be very simple. Add one for every up vote and subtract one for every down vote you get from other people on anything.


Simple answer - Submissions are more important than Comments for News.YC ;-).

If comments were given more weight tending towards that of submissions, people would have trouble moving onto the next submission, in the process of accumulating karma. I am sure YC wants people to add value through new/novel submissions.


I think aston means that he thinks he gets a full point for his comment, Not what the justification of the 0.5 factor is.

I wasn't sure on this, but I didn't think it was straight 1 point so the 0.5 might be right, I am sure other people have more evidence.


Sorry about that aston and immad. Even I believe its not 0.5 deterministically, but I think it averages to 0.5. The system appears far more adaptive/dynamic than a completely deterministic model as suggested by the formula. But again, its a model.

My friend, up-voted me once each for two separate comments, but my Karma remained the same. As a matter of fact, I observed something similar with the submissions as well. This could imply the following -

1. Karma calculation is not interrupt driven, but its timeout based (I probably hadn't waited long enough!) OR

2. The weight for the comments (/submissions?) could be a function of user Karma, with an average of 0.5 (I probably did wait long enough!) OR

3. My friend's up-voting was suspicious (Highly doubt it!!)

My bet is on 1 and/or 3.


How'd your friend up vote the same comment twice? As far as I know, the system only counts one vote per user per comment or post.


Errata - My friend up-voted me on two different comments, once each. My score should have increased by 1 Karma point going by the 0.5 weight for comments, but it did not, immediately.


The system seems to work well and I see no reason to make it more complicated. In the ideal world you might get more karma for stories you've written yourself as opposed to stories you found elsewhere and submitted, or perhaps there might by a leader board for the highest average submission rating, but the current system seems to get the job done.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: