Forum Moderators: open
That, of course, may have changed. But I believe that most of us assume that internal PR is recycled.
I thought this was impossible unless some of the inner pages had their own incoming links that could pass this individual PR back to the home page.
Perpetual PR :)
What evidence do you see to the contrary?
That's how it struck me UKG, but if Mohamed_E's point is correct......
So if your theory is right, you don't need incoming links except for the SEs to find you, then with a mass of internal links your PR should hit the roof.
I'm thinking that there has to be some lower value placed on links coming from the same site. Otherwise PR would only be a rough indication of the number of pages on a site.
My guess is that Google's current algos recognise domains/sites as entitites as well as pages and therefore whilst having a thousand pages of content and good internal links will be beneficial, external links are probably essential.
It is probably possible to move internal PR around through links but creating it is a different matter altogether.
Before trying to manipulate the PR algos, bear in mind the values reported by the toolbar may not be those used to sort search results. Google may be the David Blaine of search engines.
Kaled.
That's how I saw it, but many of the PR channeling/funneling advocates imply a (eg) PR6 can be created from a PR4 with sufficient recycled PR.
Imagine writign a site, on a single domain, that copies ALL pages on the net. Then deleting all other domains. Do you think this site of over 4 billion pages, includign all of yahoo for example would suddenly be all PR0?
Of course PR goes up with number of pages (I have sites that have high PR purely because of the volume of pages). Just that the intrinsic PR of a single page is so low that you will not notice it unless it's in the 10s of 1000s. And even then, a single incoming PR4 links is probably worth more (actually not really, more like a PR5 or 6 link).
Think about what makes sense. There is an argument for givign more import to links from a different domain of course, but PR has to come from somewhere initially, and claiming it comes from other sites simply shifts the issue, not answers it.
SN
*It is probably possible to move internal PR around through links but creating it is a different matter altogether*That's how I saw it, but many of the PR channeling/funneling advocates imply a (eg) PR6 can be created from a PR4 with sufficient recycled PR.
Well, my understanding of PR is that a page has PR1 by default, (if Google can see it of course). The PageRank algorithm migrates an amount of that out of the page via outbound links, and accumulates more via inbound links.
By creating a Page, you *do* create PR, but not much. Funnelling somewhere useful is the name of the game.
By that strategy, a PR6 page has become PR6 from PR1, let alone PR4.
If one accepts the model that the entire Internet is a single site, on different domains, linked in a complex but entirely well-defined way, then it follows that PR6 *can* be created from PR1 just be moving PR about.
I studied the "Page Rank explained" or "uncovered" or whatever it was paper (sorry - it's on a different machine from the one I'm on right now), and was moved to build some spreadsheets to iterate round some (albeit closed) systems; the effects of different linking strategies can easily be seen. Certainly, a closed system of 60 pages all of rank 1 (that's rank on a linear scale, not a log one), can be "wired up" to migrate a noticeable amount of PR up to a nominated page. And that's what I've done to one of my sites and I'm very pleased with the result.
Whilst the PR algorithm may have been tweaked since those early days, it was definitely site-independent and treated any page as a page with some in and some out links. That was the beauty of the algorithm.
And that explains why you can't make a site with PR6 that's totally closed off from the outside world - Google will drop it from its index - it's not part of "The Internet"
When GoogleGuy gave the hint "add content" I took it to mean indexable content AND more pages. That's what I did, and my rankings certainly improved - more pages where *I* control where the PR flows means more PR that I can inject into the pages *I* want.
DerekH
With linking to 400 own pages gives 0
because all get just 1/400 PR5...and
PR0 pages are PR ignored...
Why it still works with high PR pages that
have 100 000 internal pages has to do with
Google not giving PR0 to a not linked page
as it seems when the main page is high PR....
All Imho :)
However, I doubt that cross-links have created any PR. The number of pages will have created some PR (probably) but not very much.
Kaled.
PS
The page with zero external backlinks is for software currently in beta - it's not an oversight on my part.
Yes, you can recycle PR this way. This isn't a "perpetual PR" machine. What happens is if all the internal pages link back to the home page, you boost home page PR at the expense of other internal pages that could possibly be linked to. Thus, linking all the internal pages to the home page, while it raises the home page PR, squanders the opportunity to "spread the PR around more evenly" throughout the site. Recycling the PR to the home page tends to be bad strategy unless the site is highly focused, and targeting only a few keywords. Thus, doing this makes sense if you sell only fuzzy, blue widgets. However, if you not only sell all kinds of widgets, but also doodads, doohickeys, thingamabobs and a lot of other products, no way of optimizing the home page to do well on all those search terms.
It seems to me that thinking too deeply about PR is probably bad for your mental health. No-one beyond the inner circles at Google is likely to be privvy to this knowledge. Plenty of people have made guesses based on observations and deductive reasoning (myself included) but that is all they'll ever be - educated guesses.
Some people have actually read the two papers in which Page Rank is described: The Page Rank Citation Index: Bringing Order to the Web [dbpubs.stanford.edu] and The Anatomy of a Large-Scale Hypertextual Web Search Engine [www7.scu.edu.au].
I am always careful to stress that these papers were written a few years ago ( a geological age in the fast moving world of the web :) ) and may well no longer represent exactly what Google does.
I was under the impression that PR passed to inner pages cannot be "recycled" to boost the home page PR unless there are some outside links to inner pages.
I think from pure theory point of view i would agree with Mohamed_E 's assumptions. At least that is what my understanding of the original Pagerank papers are :)
From what I see PR means nothing anymore unless you got old PR.
I understand the concept of "old money" :) but must say that I fail to understand what "old PR" could conceivably be. Please clarify.
In theory, Google could assign greater weight to links a page has had for years than links that were recent. However, I'm not sure this would be a good idea. In particular, on a topic that rapidly is changing, old sites that hadn't been updated in years would rank higher than new sites that are very up to date. This would be a Bad Thing.
Sergey Brin and Larry Page thought about Perpetual PageRank machines before Google was born.
From 2.4 of "Bringing Order to the Web":
There is a small problem with this simplified ranking function. Consider two web pages that point to each other but to no other page. And suppose there is some web page which points to one of them. Then, during iteration, this loop will accumulate rank but never distribute any rank (since there are no outedges). The loop forms a sort of trap which we call a rank sink. To overcome this problem of rank sinks, we introduce a rank source [ ... ]
The paper basically tells us that the rank source effects a decay due to the normalisation at each iteration.
So, lets go with the decay factor and switch to 2.1.1 of "The Anatomy of a Search Engine":
We assume page A has pages T1...Tn which point to it (i.e., are citations). The parameter d is a damping factor which can be set between 0 and 1. We usually set d to 0.85. There are more details about d in the next section. Also C(A) is defined as the number of links going out of page A. The PageRank of a page A is given as follows:PR(A) = (1-d) + d (PR(T1)/C(T1) + ... + PR(Tn)/C(Tn))
We'll assume that d is 0.85 (I think it's higher now). We'll work an ideal feedback loop for a case where PageA receives PageRank from outside our system, large enough to make the (1-d) 'rank source' irrelevant. PageA links only to PageB, which links only to PageA.
Iterating PageRank...
PageA gets (Incoming + (PageB * d)); PageB gets (PageA * d)
PageA gets (Incoming + (PageB * d)); PageB gets (PageA * d)
PageA gets (Incoming + (PageB * d)); PageB gets (PageA * d)
PageA gets (Incoming + (PageB * d)); PageB gets (PageA * d)
etc.
Due to (d), the PR values gradually converge.
With d = 0.85, PageA ends up with about 3.6 times the incoming PR. On the Toolbar Scale, this would only be about half a notch of PR. The future for perpetual PageRank machines is not good.
In this case PageB ends up with roughly the same PageRank as PageA, but in the normal case PageB is actually a bunch of pages, with perhaps a third tier. It's not hard to see why the typical site has a PR(n) home page, PR(n-1) next level and PR(n-2) next level, etc.
Using this tool, it is easy to demonstrate the effect on the index page by back linking from all or selected lower level pages.
However I cannot say whether the algorithm is that currently used by Google.
Harry
There is a facility at WebWorkshop where you can model a web site and by changing linking strategies see the effect on the individual pages.
I certainly would love to be able to model exactly what my current link structure does, and compare that against what I want it to do.
I have, however, one problem with this approach (modelling or simulation), and it is a question of scale. I very much hope that ciml will enlighten me.
I believe that the enormous size of the web is an intrinsic part of its nature, not just a complexity that may be "modelled out". Assuming the web has one billion (US) pages (I know it is several times larger) and that my site has 1,000 pages. Then scaling down at a 1:1000 scale will still leave me with a model web of one million pages (more than I can deal with :) ) and will squish my site down to a single page, hardly enough for studying its internal structure :(
In the absence of a quantitative model I used my intuitive unerstanding of what I had read plus a healthy dose of common sense. The combination does seem to have distributed PR internally in the rather complex way I chose. It would certainly be nice to be able to confirm it.
This is a bad analogy for several reasons, none of which are that I don't own a boat. Still, I hope it helps demonstrate the point about considering the flow into, around and from your pages without needing to consider the Web at large.
I think that the basis of the whole thing is the starting points.
The whole point of a converging algorithm is that the starting points don't matter at all, surely?
That's what the more complex papers state, and what many of the illustrative papers show.
DerekH
You would think that all of the sub pages would then have the same PR but they do not. Some have pr4 and some have pr5. Strangely enough, some of the pages that the home page DOES NOT link too are the PR5 pages.
This confuses me. The home page is pr5 and is the only page with external incoming links (about 80.) You would think that pages that the home page links to would have higher PR than pages that it does not. All of the sub-pages have only internal links.
Is every page linking to every other other page is a bad idea?
Currently only about 163-167 show as backlinks depending on the particular page. It has went from the teens to over 200 depending on the month.
