Wikipedia talk:Does Wikipedia traffic obey Zipf's law?

From Wikipedia, the free encyclopedia

Rank frequency plots have the rank on the Y AXIS! —Preceding unsigned comment added by 128.153.23.240 (talk) 20:20, 8 September 2007 (UTC)

When I hit "reload" on this page, I see a rectangle in which presumable an image is to appear. But when it's finished reloading, there's not even any vertical blank space between the paragraph above the image and the one below. Michael Hardy 19:46, 16 October 2006 (UTC)

... and now I've clicked on the image and viewed it independently of the project page, and after that, when I hit "reload", the image appears normally. Michael Hardy 19:50, 16 October 2006 (UTC)
... and now it's back to its original abnormal behavior. Michael Hardy 19:53, 16 October 2006 (UTC)

Contents

[edit] approximation is way off

the 80 hits a day approximation for the least popular page is way off.. just extrapolate that line in the graph one more box to rank 10000, then another to 100,000, then another to 1,000,000, and you're well below 10 hits a day, and getting closer to 1 hit a day. somebody muddled up their maths. 81.168.22.81 00:26, 27 December 2006 (UTC)

The "hits" figure on the graph is scaled by an arbitrary constant amount, as described in the article intro: the calculation in the article re-scales it based using a single data point with a known rate to provide an eatimate of absolute rates. -- The Anome 02:51, 3 February 2007 (UTC)

[edit] Testing the predictions in this article

While we can't identify the least popular article in wikipedia, could we test this law by selecting a 100 or so articles at random, and see if the number of hits to the median of the 100 matches the Zipf's law prediction? Andjam 02:52, 21 February 2007 (UTC)

[edit] Long tail

Will Zipfs law work if the data have a long tail? futurebird 03:06, 15 March 2007 (UTC) Oh yeah, that works fine. ... never mind. futurebird 03:07, 15 March 2007 (UTC)

[edit] Other wikis follow Zipf's law

Log-log graph of page traffic in other wikis
Log-log graph of page traffic in other wikis

I had a look at whether other wikis obey Zipf's law (and not just for the top 1000 articles) using the "Special:Popularpages" function in wikis that use a mediawiki engine but aren't hosted by wikimeda or wikia. It seems like they do, and to a greater extent than I expected. In general, the greater the traffic, the more closely it obeys Zipf's law. The main exception is that the limited amount of articles (in all wikis) lead to a "drooping tail" at the end. Andjam 01:53, 9 April 2007 (UTC)