Comment score using Wilson score interval

[nremond]

Hi,

I think we can easily improve the computation of the comment score. It should be based on the Wilson score interval as explain by Evan Miller here : http://www.evanmiller.org/how-not-to-sort-by-average-rating.html

Thus, I would suggest to patch your compute_score_comment() method as follow :
https://gist.github.com/1343554

What do you think ?

[jellomaster]

Yes I think this is a good idea.

[nremond]

About the news sorting, I found this interesting blog entry : http://blog.linkibol.com/2010/05/07/how-to-build-a-popularity-algorithm-you-can-be-proud-of/

The formula here : http://blog.linkibol.com/resource/linkibol.PNG could correspond to what antirez is expecting when he put this comment in the code :
# FIXME: For now we are doing a naive sum of votes, without time-based filtering, nor IP filtering.

[antirez]

This change looks great! I'll merge it in the next days. Thanks.