This subject interests me very much but I realize that it is considered trivia by others - that's cool. Dismiss and move on or read and leave comment as you see fit. Different strokes for different folks.

I was prompted to prepare this blog after reading a forum post this morning: Forum:Most_visited_pages

Whether my goals are the same as the author of that forum post or not I do not know for sure but for the sake of comparison you may care to take a close look at w:c:vsk:Category:Members#Manage_your_Favorites and the references provided there - starting with w:c:vsk:Template:Favorites. (Note the wiki must have the extensions: Semantic Mediawiki and Semantic Forms enabled to make use of the Manage your favorites feature described at that link.)

I was in search of a measure of popularity for the subject matter of certain wiki pages. Initially this was driven by the desire to rate the popularity of certain community-member created and uploaded files used in that game but I later realized that the solution was able to be applied to any wiki article.

  • The subject of an article may be popular even though the wiki article itself is poorly written (especially if there is no descriptive text written!)
  • An article may be very well written but it may be the case that few or no readers are interested in the subject of the article!
  • I assume that the 5-star-voting system at Wikia is there to capture opinion about the quality of a wiki article. If this is a correct premise then it begs several questions:
    1. Do past votes get purged as soon as the article is subsequently edited?
    2. Do past voters get notified that a page they voted on has been edited?
    3. Do past votes get purged after some period of time has passed?
    4. Suppose the subject of the article gets updated (maybe it is a game wiki where that game is subject to revision so the onus is on the wiki to keep current) and yet the wiki article is never updated. How can past votes remain relevant?
    • Indeed I know of wikis where that 5-star-voting link is hidden by sysops because they feel it is rubbish for some (or all) of the reasons hinted at by the questions listed above.

I was searching for a somewhat democratic, one-user/one-vote measure of popularity.

  1. The 5-star system tends to be skewed by users who vote high since users who are not impressed with an article either do not bother to vote at all and when they do decide to express their displeasure the lowest vote they can give an article is a 1-star as opposed to a zero-star.
    • lol - it would be a very different matter if a page that is visited by a registered user who does not bother to vote is then automatically given a zero-star rating associated with that user!
  2. A user ought to be able to see how they voted at a page and be able to change their vote at any time
    • Indeed the 5-star voting system accommodates this
  3. It ought to be easy for a user to locate pages that they have voted on in the past. Ideally I could run such queries as:
    • show me pages I gave 3 or more stars to in the past
    • show me pages that I voted on over 90 days ago
    • show me pages that I gave 2 or fewer stars AND which have been modified since I voted
    • ... and so on and so forth.
  4. A vote should never expire or be purged but a user ought to be able to withdraw their vote at any time
    • As best I can understand the 5-star voting system does accommodates this but it sure would be nice to be able to leverage one of the query examples listed above and say:
    • remove my vote for each page that I voted on over 90 days ago AND which has been modified since I voted
  5. Other users ought to benefit from knowing both the number of votes a page gets as well as the size of that voting/abstaining population.
    • I believe this to be a major shortfall with each one of the candidate "popularity metrics" listed below.

How many different ways are there to judge popularity?
If you take inventory of the tools available to you there are many "similar but different" measures to choose from:
  1. Special:Mostvisitedpages : as has been noted elsewhere this counts multiple page visits (indeed even page refreshes are counted) by the same user and so is very easy to skew/bias the result
  2. Special:Mostpopulararticles : I cannot understand this one - it does not match the "highest voted" list avalable via the widget: Top content
  3. Special:MostLinkedPages : maybe useful to architects of a wiki but not so much use to a casual reader/browser
  4. Special:MostCategories : maybe useful to architects of a wiki but not so much use to a casual reader/browser
  5. Special:Watchlist : of immediate interest to a registered user and heavily customizable via:
  6. Widget - Bookmark : useful to the registered user but there is no way to see how many users have bookmarked a given page, so that measure of popularity is lost
  7. Widget - Top content : really 4 lists in the one widget with 2 out of 4 being some form of popularity metric:
    • Featured : this is a list specified by the sysops of a wiki - useful for featuring important articles
    • Most visited : presumably this is derived form the same data as Special:Mostvisitedpages since the two lists do match one another
    • Highest voted : presumably this one is based on input provided via the 5-star-voting system - note that this list does not match the one at Special:Mostpopulararticles
    • Newly edited : I cannot figure understand this one - it does not match the Special:Recentchanges even when the latter is filtered to show only articles in the Main namespace
  8. Widget - Most visited : identical to Widget - Top content/Most visited
  9. Widget - Watchlist : vaguely resembles Special:Watchlist so it is possible that the few discrepancies can be explained by the details given at Help:Watchlist#How does it work?
  10. ... did I miss any others?

Going back to the enumerated requirements of my sought-after one-user/one-vote measure of popularity it is easy to see that none of the available tools satisfy those numbered requirements.

The solution used at w:c:vsk is probably not perfect either. However it does achieve the following:

  1. one vote per user
  2. equal weighting of votes
  3. the total population of potential voters (#members at Category:Members) is based on that list of users who have made at least one edit at the wiki
    • users may voluntarily remove their username from this "total population" using a pair of comment delimiters at their user page
    • This leverages the bot account named Wikia and the MediaWiki:Welcome-user-page and was a compromise because there is no way of knowing how many unregistered readers (or registered readers who never make an edit) ought to be included in that population.
  4. users can see and change their vote at any time
  5. otherwise votes never expire and never get automatically purged
  6. users may display list of pages that they voted on at their user page (comma separated list and table of results are possible).
  7. queries like those listed above are possible using semantic mediawiki inline queries
  8. at the discression of the page author, other visitors to the page may be shown a simple popularity index: #fans/#members
  9. while vote tampering is possible such vandalism is just as easy to undo as rolling back the page edits by a specific user
  10. can be implemented without special support from Wikia staff

-- najevi 02:19, December 1, 2009 (UTC)