On 5/3/05, Andrew Dunbar <hippytrail(a)gmail.com> wrote:
Rather than simply most views, it might be a better
idea to show which
articles are gaining the most popularity compared to their usual level
of activity and which are losing popularity. This will prevent
advertising the Hitler and Evolution type of articles. Google
Zeitgeist also does something like this.
Still, that would need some more refinement. First, is the comparison
done by substraction or by division? In the first case, you will still
get the high-base-level pages (such as [[Wikipedia:Village pump]] if
you go beyond articles) pop up often; in the latter case it is the
low-base-level pages that can pop up without having anything special
("This article was created 1 year ago, and hasn't been edited since,
but now it has had a spelling correction - that's 52 times its usual
level"). Then there is the problem of people incrementally editing a
page, giving it many edits. This could be solved by counting the same
editor twice only if there is another editor in between. We could even
count the number of editors sec. The advantage of that is that an edit
war is 'cooled down', counting only as 2 editors, the disadvantage
that the number might be quite low even for hot pages. Finally, there
is the question of what to do with new articles - Do we define some
kind of 'base level for new articles', or are they simply disregarded
for Zeitgeist purposes?
Perhaps the best is a formula that counts in various factors (basic
number of edits, number of 'edit strings', number of editors, time
between edits (if something is edited twice in a short time, the
second edit is more likely to be caused by the first one directly
rather than both be caused by a 'Zeitgeist'), non-reversion edits,
etcetera). Disadvantage of that would be that base levels are hard to
calculate.
Andre Engels