Nearly two years ago, as part of my ongoing Metafilter Statistics project, I decided to see which websites the Metafilter community was linking to most frequently. With Matt Haughey’s help, I analyzed a complete dump of every post and came up with this list.
Now, two years later, I ran it again. Some of the results are surprising, as you can see below.
In the space of two years, YouTube launched and became the 15th most popular site, Flickr jumped 2,400 places to #14, and Wikipedia went from #66 to #2. Other big winners included Ask Metafilter, Times Online, Archive.org, and MSNBC. Huge losers that fell off the Top 50 entirely were Nandotimes.com (stopped publishing), News.cnet.com (changed domains), Observer.co.uk, and ZDnet.com.
Update: Some people in the Metatalk thread had questions about my methodology. The number next to each link is the jump in rank from the 2002 list. I’m only counting links in the front-page post itself, not the comments. And in the complete list, I’m excluding the 26,123 sources with only one link to conserve space.
Top 50 Metafilter Sources: 1999-2006 (View all…)
- metafilter.com
- en.wikipedia.org
64
- news.bbc.co.uk
1
- nytimes.com
1
- amazon.com
6
- washingtonpost.com
- cnn.com
3
- imdb.com
38
- geocities.com
4
- guardian.co.uk
3
- google.com
5
- bbc.co.uk
8
- pbs.org
6
- flickr.com
2406
- youtube.com NEW!
- us.imdb.com
1
- sfgate.com
3
- wired.com
6
- story.news.yahoo.com
11
- salon.com
11
- dailynews.yahoo.com
16
- npr.org
16
- msnbc.msn.com
83
- cbc.ca
5
- boston.com
7
- msnbc.com
16
- abcnews.go.com
10
- memory.loc.gov
21
- abc.net.au
24
- livejournal.com
52
- members.aol.com
- usatoday.com
4
- latimes.com
11
- whitehouse.gov
6
- news.yahoo.com
23
- time.com
9
- slate.msn.com
16
- newyorker.com
4
- angelfire.com
4
- newscientist.com
17
- forbes.com
21
- theregister.co.uk
18
- telegraph.co.uk
9
- cbsnews.com
23
- cgi.ebay.com
10
- villagevoice.com
13
- web.archive.org
39
- apple.com
12
- news.google.com
51
- loc.gov
57
Only ~3 blogs in the top 100. (And a formatting suggestion: make the color of the rank change number match that of the arrow, easier to scan for the ups/downs that way.)
This is really wonderful. I appreciate having it very much.
Is it too ungrateful to ask for a little clarification? You say that Wikipedia went from #64 to #2, but actually it jumped 64 places. That’s what all the numbers mean, right, the change in rank?
This analysis is of *all* links on Metafilter, right? Comments and everything? Is it restricted to http://www.metafilter.com, or does include the other subdomains?
Also, “http:”, home of dead links everywhere, is #137!
Links to Google’s cache are prominent, with “66.102.7.104”, at #158, “72.14.207.104” at #203, “64.233.161.104” at #356, “72.14.203.104” at #473, and “216.239.57.104” at #498.
“195.2.85.140” at #212, is a massive store of 23,000+ pictures of cigarette packs. I imagine there must have been a FPP about the site, and it was linked to a shitload in the comments. Can’t seem to find such a post though.
jamjam: Sorry, that was a typo on my part… Wikipedia went from #66 to #2, which is a 64 place jump in rank.
blasdelf: the post is here
Shouldn’t the results from imdb.com and us.imdb.com be merged?
Why is loc.gov listed separately from memory.loc.gov? The Memory project is just a fancy interface.
They’re listed separately because the Metafilter community used two different domains. I can’t just automatically merge all subdomains together, because it will eliminate a bunch of valid sites hosted on subdomains, like Typepad and Blogspot blogs. Many sites, like Yahoo and Cnet, also use subdomains for their different properties.
I’m not going to manually comb through tens of thousands of domains to merge some and not others.
I understand the problem, but the results as presented have a major bias. It suggests that these results are little too arbitrary, like comparing raw scores from the LSAT and GREs. Or something. Plus, if it turns out that more people link to various blogger blogs than wikipedia entries, we’re missing out on some really good GYOBFW snark. 🙂