archEngineGeek.com/graphs/dey01.gif

http://www.SearchEngineGeek.com/graphs/deg01.gif

(Note to webmasters: Feel free to hot link to the above graphs or copy them to your own site if you publish this report. Feel free to remove this note as well.)

The number of links were grouped in this way in order to increase the number of data points available. Unfortunately it also reduces the precision of the results. One is able to see that 91-100 link

Lots of research has focused on inbound links to a site, but little has focused on the number of links actually on a page (outbound or to other parts of a site). Many SEO gurus have recently been talking about something they call "PR Leak" which seems to be a theory that the more outbound links you have, the more your page rank on Google "leaks" away. That concept isn't found in the academic papers published by the founders of Google, but does seem to be accepted by a majority of SEOs. I decided it was time to take a look at the number of links present on a page and how that number correlates with ranking.

The methodology: I gathered the results of the queries that were naturally performed last month by myself and three associates using the two leading search engines and analyzed them. I counted the number of links on the page (references to "href") and tabulated the results against the ranking of the URL in the search results. The tabulated results were finally converted into a normalized "ranking correlation." The results for each of the two leading search engines were kept separate so that we could discover any differences between the two leading search engines for this factor.

The resulting graphs show the results for groupings of number of links normalized into a number between -100 and +100 showing the likelihood of being ranked higher/lower. A value of +100 shows that all 10 rankings were in the proper order to show that pages of the studied value ALWAYS rank HIGHER than pages of another value. A value of -100 shows that all 10 rankings were in the proper order to show that pages of the studied value ALWAYS rank LOWER than pages of another value. Numbers in between show the varying likelihood of rankings proportionally between -100 and +100.

That is the number you see on the Y-axis. On the X-axis, we have the number of links found. They are grouped into sets of 10 in order to increase the statistical significance with the amount of data we had available to analyze. Here are the graphs for the two leading search engines:

http://www.SearchEngineGeek.com/graphs/dey01.gif

http://www.SearchEngineGeek.com/graphs/deg01.gif

(Note to webmasters: Feel free to hot link to the above graphs or copy them to your own site if you publish this report. Feel free to remove this note as well.)

The number of links were grouped in this way in order to increase the number of data points available. Unfortunately it also reduces the precision of the results. One is able to see that 91-100 links

f links present on a page and how that number correlates with ranking.

The methodology: I gathered the results of the queries that were naturally performed last month by myself and three associates using the two leading search engines and analyzed them. I counted the number of links on the page (references to "href") and tabulated the results against the ranking of the URL in the search results. The tabulated results were finally converted into a normalized "ranking correlation." The results for each of the two leading search engines were kept separate so that we could discover any differences between the two leading search engines for this factor.

The resulting graphs show the results for groupings of number of links normalized into a number between -100 and +100 showing the likelihood of being ranked higher/lower. A value of +100 shows that all 10 rankings were in the proper order to show that pages of the studied value ALWAYS rank HIGHER than pages of another value. A value of -100 shows that all 10 rankings were in the proper order to show that pages of the studied value ALWAYS rank LOWER than pages of another value. Numbers in between show the varying likelihood of rankings proportionally between -100 and +100.

That is the number you see on the Y-axis. On the X-axis, we have the number of links found. They are grouped into sets of 10 in order to increase the statistical significance with the amount of data we had available to analyze. Here are the graphs for the two leading search engines:

http://www.SearchEngineGeek.com/graphs/dey01.gif

http://www.SearchEngineGeek.com/graphs/deg01.gif

(Note to webmasters: Feel free to hot link to the above graphs or copy them to your own site if you publish this report. Feel free to remove this note as well.)

The number of links were grouped in this way in order to increase the number of data points available. Unfortunately it also reduces the precision of the results. One is able to see that 91-100 link

s were kept separate so that we could discover any differences between the two leading search engines for this factor.

The resulting graphs show the results for groupings of number of links normalized into a number between -100 and +100 showing the likelihood of being ranked higher/lower. A value of +100 shows that all 10 rankings were in the proper order to show that pages of the studied value ALWAYS rank HIGHER than pages of another value. A value of -100 shows that all 10 rankings were in the proper order to show that pages of the studied value ALWAYS rank LOWER than pages of another value. Numbers in between show the varying likelihood of rankings proportionally between -100 and +100.

That is the number you see on the Y-axis. On the X-axis, we have the number of links found. They are grouped into sets of 10 in order to increase the statistical significance with the amount of data we had available to analyze. Here are the graphs for the two leading search engines:

http://www.SearchEngineGeek.com/graphs/dey01.gif

http://www.SearchEngineGeek.com/graphs/deg01.gif

(Note to webmasters: Feel free to hot link to the above graphs or copy them to your own site if you publish this report. Feel free to remove this note as well.)

The number of links were grouped in this way in order to increase the number of data points available. Unfortunately it also reduces the precision of the results. One is able to see that 91-100 link

the studied value ALWAYS rank LOWER than pages of another value. Numbers in between show the varying likelihood of rankings proportionally between -100 and +100.

That is the number you see on the Y-axis. On the X-axis, we have the number of links found. They are grouped into sets of 10 in order to increase the statistical significance with the amount of data we had available to analyze. Here are the graphs for the two leading search engines:

http://www.SearchEngineGeek.com/graphs/dey01.gif

http://www.SearchEngineGeek.com/graphs/deg01.gif

(Note to webmasters: Feel free to hot link to the above graphs or copy them to your own site if you publish this report. Feel free to remove this note as well.)

The number of links were grouped in this way in order to increase the number of data points available. Unfortunately it also reduces the precision of the results. One is able to see that 91-100 link

archEngineGeek.com/graphs/dey01.gif

http://www.SearchEngineGeek.com/graphs/deg01.gif

(Note to webmasters: Feel free to hot link to the above graphs or copy them to your own site if you publish this report. Feel free to remove this note as well.)

The number of links were grouped in this way in order to increase the number of data points available. Unfortunately it also reduces the precision of the results. One is able to see that 91-100 links rank much higher than 1-10 links, but you are unable to see if 77 links rank differently than 79 links (for example).

The result is very conclusive. Both leading search engines rank pages with more links much higher than pages with fewer links! Once again, it appears that the SEOs touting the "PR Leak" theory are simply wrong. If their theory held any weight at all, we should see the exact opposite. Pages with more links should rank lower on average.

Notes:

1. There was no exercise to attempt to isolate different keywords. I merely took a random sampling of the queries performed by myself and three associates during the month.

2. This is merely a correlation study, so it cannot be determined from this study whether the leading search engines purposefully entertain this factor or not. The actual factors used may be far distant from the factor we studied.

Copyright 2004 Jon Ricerca