Re:How Not to Sort by Average Rating (2009)
by Houshalter on Thursday July 09, 2015 @06:33AM
(#9857096)
There is a much simpler and elegant method. Just rank posts by their probability of getting an upvote. This is just (upvotes+1)/(upvotes+downvotess+2).
[ Parent
| on HN ]
Re:How Not to Sort by Average Rating (2009)
by stdbrouw on Thursday July 09, 2015 @06:57AM
(#9857162)
This gives an advantage to new posts for which the probability is much more uncertain: it's easier to get 1 upvote and 0 downvotes (rank 2/3) than to get 1999 upvotes and 999 downvotes (also rank 2/3). Maybe that's what you want, but the post is exactly about those cases when this is not what you want.
Re:How Not to Sort by Average Rating (2009)
by dredmorbius on Thursday July 09, 2015 @12:54PM
(#9859114)
New posts frequently start with a disadvantage though in existing systems. Temporarily biasing them favourably increases odds of any moderation. Alternatively you could present them only to a subset of readership. I've suggested this as a solution to HN's new submissions queue problem.
Increase the presentation as ratings increase.
Re:How Not to Sort by Average Rating (2009)
by Houshalter on Thursday July 09, 2015 @03:29PM
(#9860164)
You can of course use a better prior if you don't think new posts are 50% likely to receive an upvote next. Otherwise I don't see how that behavior is undesirable or against the goal of the article.
Re:How Not to Sort by Average Rating (2009)
by fludlight on Thursday July 16, 2015 @11:42AM
(#9897884)
How about fn(upvotes, downvotes)/fn(pageviews, prominence)?
Re:How Not to Sort by Average Rating (2009)
by anon4327733 on Thursday July 09, 2015 @02:55PM
(#9859969)
That is true for a uniform prior but not a general prior.