Re:How I ended up with so much Hacker News karma
by gjm11 on Monday December 03, 2012 @09:06AM
(#4865380)
More than that. [EDIT: When I wrote this, jgrahamc's comment said only "IIRC the HN algorithm removes outliers". Since then he's added much more. But I don't believe in deleting comments merely because edits to their parents have made them less valuable.] The "average" in the HN profile is something along the following lines: take the 6th to 25th most recent comments, throw away the one with most points, and take the average of the others. So it ignores old comments, it ignores very recent comments, and it ignores some kinds of outlier. Oh, and it's only actually recalculated every few days.
Personally I don't much like this: average karma is a fairly good proxy for what I want to be optimizing in my HN participation, but the way HN reports it more or less pessimizes it as a feedback tool for me. (I'm a fairly infrequent and "bursty" commenter; HN's reporting of average karma isn't up-to-date enough to tell me what HN thinks of what I've been doing recently, nor is it long-term enough to tell me what HN thinks of me overall.)
The right answer, of course, is not to care much about karma. This is made easier by the fact that the correlation between HN karma and my own judgement of comment quality is weak, and I'm arrogant enough to think the latter is more accurate.
Personally I don't much like this: average karma is a fairly good proxy for what I want to be optimizing in my HN participation, but the way HN reports it more or less pessimizes it as a feedback tool for me. (I'm a fairly infrequent and "bursty" commenter; HN's reporting of average karma isn't up-to-date enough to tell me what HN thinks of what I've been doing recently, nor is it long-term enough to tell me what HN thinks of me overall.)
The right answer, of course, is not to care much about karma. This is made easier by the fact that the correlation between HN karma and my own judgement of comment quality is weak, and I'm arrogant enough to think the latter is more accurate.