bsitu wrote:

This looks like a bug for AFT?

Guh. Muscle memory. Sorry!

The problem was probably introduced when splitting featured & resolved: https://bugzilla.wikimedia.org/show_bug.cgi?id=38818

Now, upon resolving feedback, that feedback's relevance score gets a +5 increase. This would have to be turned into a -5, making the feedback _less_ relevant.

Pushed fix to Gerrit (https://gerrit.wikimedia.org/r/#/c/21318/) & prototype

Thanks! :)

Resolved feedback is appearing below featured, but above everything else. Unless feedback is hidden, junked or requested oversight, resolved should be *below it*. This is still not fixed,

Taking the first one as example
(http://en.wikipedia.org/wiki/Special:ArticleFeedbackv5/Higgs_boson/150728),
the reason appears to be the large amount of "helpful" votes (106-23=83), which
means +83 on relevance score, making the score unusually high.
This keeps it at the top of the relevance-score based filter (which is the one
anon users see by default) which makes it even more likely for users to mark it
as helpful, ...
You get the picture.
Maybe we should only display feedback within the last month for the relevant
filter (not yet sure how I'll implement this once sharded though, so other
ideas are welcome as well)
Or we make sure that only a certain amount of positive actions count to the
relevance score, a number that it just a bit under the negative score that
hiding/resolving packs?

Also: it might help to have marking something as resolved also automatically
unfeature the feedback post if it was previously featured (because it now keeps
the 50 points it received upon featuring)
Had this been done, the feedback had probably never been up there for so long
and it wouldn't have had the chance to run away from other articles.

That sounds sensible. Any way we can have "if you hit resolves, voids all other markins and adds -10" or whatever?

• Bug 40537 has been marked as a duplicate of this bug. ***

Pushed to Gerrit (https://gerrit.wikimedia.org/r/#/c/25498/) & prototype