Forum Moderators: Robert Charlton & goodroi

Message Too Old, No Replies

Flaw in the Page Rank algorithm?

         

tedster

8:49 pm on Dec 27, 2005 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



I found this interesting pdf:

Faults of PageRank [www2.mta.ac.il]

The author observes that PageRank's functioning depends on having a "well connected" web graph, but in reality there are areas of the web that are "sinks" -- pages/nodes that are not well connected. Page and Brin knew about this in the beginning, and they made a modification to accommodate the issue. But apparently, their modifcation did not result in a true fix.

What the author sees is that these weakly connected "sink" pages can get unnaturally high PageRank. In fact, by REMOVING links into the main body of the web graph and thus creating a sink, low ranking pages can make significant jumps in ranking.

Some people have noticed that the Google toolbar reports unusually high PR (a 5 with very few IBLs, for instance) for pages in more loosely connected areas of the web. And in other more highly connected areas of the web, pages can struggle to get above a PR2 or 3. Often there's a guess that Google makes PR work differently for different areas of the web -- and that never made much sense to me. PR is not in any way dependent on content.

Now I'm thinking this mathematician may be onto an explanation.

tedster

12:46 am on Dec 28, 2005 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



Discussion here:

[webmasterworld.com...]