Proven Correlations between linking and ranking and other SEO factors.

[mikeb]

Apologies if this link has already been posted here - I did not see it - but I think this article is well worth reading

http://www.theopenalgorithm.com/correlation-data/links/

Of course the usual caveats apply. In statistical terms correlation does not imply causation.

[bwelford]

Perhaps I'm missing the obvious but I don't see what I learn that is new from this. Google has said that links are part of the set of factors that influence ranking. This confirms that. ... and so?

[mikeb]

Perhaps I'm missing the obvious but I don't see what I learn that is new from this. Google has said that links are part of the set of factors that influence ranking. This confirms that. ... and so?

The significant thing is that post penguin nothing of great substance seems to have changed.

So penguin seems to be confined to details such as anchor text distribution., not the start of a dramatic move away from importance of linking.

[bwelford]

Well I'm not sure. This is all from the same people who were advocating PageRank sculpting for over a year thinking it was beneficial when the reality was that it had zero effect.

There can be two reasons why someone cannot see a change:
1. There is no change, or
2. There is a change but their testing methods are ineffective.

I rather think reason 2 is what's happening here.

[mikeb]

Well I'm not sure. This is all from the same people who were advocating PageRank sculpting for over a year thinking it was beneficial when the reality was that it had zero effect.

There can be two reasons why someone cannot see a change:
1. There is no change, or
2. There is a change but their testing methods are ineffective.

I rather think reason 2 is what's happening here.

It could well be. I agree....At this stage any evidence is worth looking at, and all of it is potentially spurious.

I still think the best penguin evidence is that from microsites and brad callen ie it is essentially about anchor text distribution

I have long believed that google would find it more or less impossible to extract themselves from the importance of linking in general, despite some commentators saying it is now more about "on page" than off, and the only way they can avoid people gaming how the system works is by allowing or helping the SEO community to mislead itself.

I cited a mindgame...if WIKI links still matter,(as a random example) then google would be dumb to say so, they would say "no follow links do not get followed" and "wikil links are no follow" - ( which is what they do say) whilst carefully avoiding saying that they do not follow them (ie robot through) but still count the name of the domain in the anchor as a signal called wiki authority - which could be a separate signal, so allowing google to say "No it does not matter" when actually meaning "it does". in the hope that the only wiki links are then natural.

[bwelford]

I have long believed that google would find it more or less impossible to extract themselves from the importance of linking in general, despite some commentators saying it is now more about "on page" than off, and the only way they can avoid people gaming how the system works is by allowing or helping the SEO community to mislead itself.

A much simpler way for Google is to stop making public in any way the PageRank value they attach to any given web page. That immediately stops all the false hype and commerce around links from high value PageRank pages.

[mikeb]

A much simpler way for Google is to stop making public in any way the PageRank value they attach to any given web page. That immediately stops all the false hype and commerce around links from high value PageRank pages.

Matt cutts has already said that even when he joined google when computers were driven by steam...the model they used was far more sophisticated than the simple "page rank" of the patent.

I think the reason they leave it there is to actively mislead - that is to make webmasters chase their own tails on collecting links that google can not only discount for other reasons but also penalize (eg penalize links from a high rank page with few incoming links and no content - typical domainers stuff)

[bwelford]

I think the reason they leave it there is to actively mislead - that is to make webmasters chase their own tails on collecting links that google can not only discount for other reasons but also penalize (eg penalize links from a high rank page with few incoming links and no content - typical domainers stuff)

I can't see how that in any way helps Google achieve whatever its business goals might be. Presumably Google is out to make money, not to create a gaming area for all those bright PhDs to play mind-games with whoever from the outside chooses to join in.

[mikeb]

I can't see how that in any way helps Google achieve whatever its business goals might be. Presumably Google is out to make money, not to create a gaming area for all those bright PhDs to play mind-games with whoever from the outside chooses to join in.

There is no surprise in what I said:

Google cannot measure the things it would like to be important - like abstract definitions of quality and an arcane definition of "natural" linking.

All it can do is measure a lot of things algorithmically, test to see how they correlate with subjective measures of guinea pig quality, and test sets of manually reviewed link patterns, then use those measurements which do correlate to decide ranking.

They know full well that if anyone knew the details they would also know how to game the system. So they have to adopt a several pronged strategy - first to conceal the truth in as far as they can, by being "economical" with the truth - second to try to avoid link spamming of things they actually use as authority (see my mind game on wiki) by actively or passively misleading on what those sources are.

Third to look for objective evidence of people trying to game the system and actively penalise gaming- which is more or less matt cutts job description.

So on the basis that googles measure of authority is far more sophisticated than page rank, the way it helps is to allow google to measure the difference between link patterns which focus primarlly on page rank against those which (they deem "natural") focus on their more sophisticated measure of authority. If there is an imbalance, they have an objective measure of "gaming" the system. Whilst forums like warrior are awash with sellers and gamers of page rank, it is in googles interests to allow those discussion to rage unabated. Which is passively rather than actively misleading on what really matters. But to do that they have to leave the metric there...

If they are leaving the numeric page rank there and visible it is because they believe it will help with better organic ranking - by getting gamers of the system to head the wrong way.

[bwelford]

If they are leaving the numeric page rank there and visible it is because they believe it will help with better organic ranking - by getting gamers of the system to head the wrong way.

... Or perhaps it is just inertia and those who would say it should be hidden haven't managed to persuade those who say stay with the tried and true. In general Google has been much more open recently about what should appear in webpages of quality. This gaming argument for displaying PageRank would be quite counter to that.

However I come back to my key question. How does this fixation on continuing to use PageRank help Google to achieve its business goals? Google knows much more about how people are looking at and clicking away from web pages. All that data is a much better resource to determine which are the quality web pages. That would clearly help them to deliver better web pages in their SERP results, which would in turn help to grow their search audience and thus their advertising revenues.