Forum Moderators: open
8.9 if there is just the one link on s
7.3 if there are twenty links on s
6.5 if there are a hundred links on s
That's assuming a damping factor of 0.85.
==============
This means that a site with only ONE inbound link coming from a page with PR 9 AND where the link is the only one on that page (theoretical), gets a PR of 8.9.
Another page on another site, with 10 inbound links from pages with all PR 8, all with ONE link on each page, will only get a PR of 7.9 or a little higher. Worse: even by adding a site having a PR of 9, but providing 10 links on it, this site still will play second rank to the first one.
I would consider the second site to be probably more valuable to visitors than the first one, assuming all other elements equal.
Is the philosophy behing the actual PR algo flawed? What do you think? Googleguy? And if it is, how can it be improved in a way that the second site (remember: all other things equal) would score better than the first in the PR?
1. Less emphasis is given to yahoo and ODP listings. I see the weight of this factor as being a stop gap (best guess of quality of the "linker") as Google's PR algo develops.
2. Google starts seriously to give higher weigtings to links from highly relevant sites, rather than just high PR sites. Sites ideally should not have PR overall, but PR for a certain topic/cateogory.
Both of these things will help decrease PR "spam" and the self-defeating rush to sell PR as a commodity.
Another page on another site, with 10 inbound links from pages with all PR 8, all with ONE link on each page, will only get a PR of 7.9 or a little higher.
That's not right. A single link from a PR 9.0 page (with no other links on it) will give a target page PR 8.9. Ten inbound links from PR 8.0 pages would give it PR 9.1.
(Still assuming the log base is 7 and the damping factor is 0.85.)
So pages that have been online a long time tend to have higher PRs, just as established organisations and older people tend to be more highly regarded and older papers cited more often. And links/PR are just like reptutation in being dependent on social, financial (cf. thread on buying PR), and personal factors as much as on any attempt to measure quality.
The way I put it to organisations is that static links to their web sites are a form of "reputation capital", and PageRank is a measure of that.
Another thought...there are 404 pages that are above mine in the listings for my search phrase. An ODP page would less likely be listed as a 404. Possible but less likely.
Google relies on Yahoo and ODP precisely becuase they are the best - but of a bad lot maybe only - stamper of authority.
My point is long term, other "authorities" - and mainly suject matter expert ones, will emerge and may be much better than yahoo and odp as "authirities" in certain subject areas.
Danny, also I agree with you. But google's key selling point has initially been Page Rank - look at the way they spin it - and they say it leads to a "quality" index. The assumption is that reputation equals quality, which to me makes absolute sense. And much more sense of course than the number of times and where a keyword appears on your page!
Even with you example of long existing sites like yours and ours, it does take repeated demonstrations of delivering to get a "reputation", just the same as in the real world! It also takes a fairly long time to lose it, and you have to screw up more than a few times once you have established a reputation.
1) PR is supposed to be a measurement of your chance of randomly arriving at a page on the web.
2) This is more or less how it is - IF YOU WERE TOTALLY BLIND.
3) People are not blind and are much more likely to click on links at the top of a page vs. the bottom. Also more likely to click on big links than small.
4) Therefore you have pages like amazon.com's that have a tiny link way at the bottom for privacy and terms of use. Same with many other sites. These aren't getting clicked on.
5) Now don't get me wrong - Google still puts more emphasis on large vs. small and such for IR measures, but not for PR itself.
6) The closer PR matches real likely hood of viewer clicking - the more accurate it will be.
Normal PR is still valuable. In fact - it is why we are all using google today and not Altavista.
However, it could be improved upon and a weighted PR would be a good measure to add in the equation.
If the logarithmic base is around 7 (which is what Chris_R estimates), then a link from a page s with PageRank 9.0 will, by itself, give the target page a rank of
I would say the log factor could be anything between 6 and 10, the problem being, it is difficult to prove what it is exactly.
Why?
1.because we do not know what dilution factor Google takes into account for the number of links on a page. Is it realy linear?
2. Does it make a difference if the links come from the home(index) page or a deeper lying page?
3. Are external links treated differently from internal links?
4. Do 20 motivated votes from a PR5 count more than one from a PR7?
I would be much obliged to see someone come up with some kind of proof taking all these factors into account. For my own site I have not been able to put it into one simple formula with an exact log rate, dilution level or dampening factor to be universal for all situations...
The original formula is rather dated, surely they have been busy since.
Chris is right we are using Google because of Pagerank, the flaw is that most people think Pagerank equals ranking... Google could be more clear in what it means and does not mean. Maybe there would be less Pagerank hype around then.
I have to differ with those that think Yahoo and ODP are good general measurements of quality. I've personally tried to get different sites of differing quality into each on numerous occassions. I've had much more success getting the lower quality pages into these "arbiters of quality." They may have been good at determining quality in the past but they are terribly slow to put up new links and the choices to include or not are frequently not evidence of intelligence.
So I've got to agree with chiyo.
The PR philosophy is right. I don't think anybody doubts about it. You can argue about the weight it gets, the way is is calculated, and so on. But the principle of trying to 'qualify' the value of a page based on a number of variables that are 'off page' is good.
Danny, I tend to agree with you. But I would appreciate it if you could explain two elements.
1. You have the talent to explain mathematical algo's in a simple way. You write:
=================
Another page on another site, with 10 inbound links from pages with all PR 8, all with ONE link on each page, will only get a PR of 7.9 or a little higher.
That's not right. A single link from a PR 9.0 page (with no other links on it) will give a target page PR 8.9. Ten inbound links from PR 8.0 pages would give it PR 9.1.
(Still assuming the log base is 7 and the damping factor is 0.85.)
=================
Could you please explain in a simple way how you get those numbers?
2. You write:
===================
Don't think about PR as a measure of quality. It's more a measure of "reputation".
So pages that have been online a long time tend to have higher PRs, just as established organisations and older people tend to be more highly regarded and older papers cited more often. And links/PR are just like reptutation in being dependent on social, financial (cf. thread on buying PR), and personal factors as much as on any attempt to measure quality.
The way I put it to organisations is that static links to their web sites are a form of "reputation capital", and PageRank is a measure of that.
==================
That is exactly what I am saying. If some site can 'buy' or 'get' ONE link from a PR9 page with only ONE outbound link, this site gets a 'reputation' beating the reputation of an 'old' site which has been building maybe hundreds of inbound links, but of lower PR value. This doesn't match with the 'reputation' theory. I don't want to go in political discussions, but who would YOU believe: ONE president who says: 'site X is OK', or 100 senators who say 'site Y is OK'?
I still think the algo (besides the weight of PR in SERP's, which is another theme) should be improved.
A single link from a PR 9.0 page (with no other links on it) will give a target page PR 8.9. Ten inbound links from PR 8.0 pages would give it PR 9.1.
pvdm asked:
Could you please explain in a simple way how you get those numbers?
Ok, it's not that tricky. PageRank is a logarithmic representation of an underlying PageValue (let's call it). PageRank 9 = PageValue 79, PageRank 5 = PageValue 75, etc.
If the damping factor is 0.85, then a page passes on that much of its PageValue, split between the links on the page. So if we have ten pages at PR 8, they can contribute a total PageValue of 10 x 78 x 0.85 = 49 000 000. Converting that back to PageRank, we get log7 49 000 000 = 9.1
With lots of assumptions...
Next week we move on to graph theory. The set text is Robin J Wilson's Introduction to Graph Theory (Addison-Wesley 1997).
That site deserves a PR8 because no one in their right mind would own a PR9 site and not link out wisely.
1/.A new site will take time to acquire links. It may be more relevant than all the other sites out there, but it might take 6 months to a year before it acquires a suitable PR for a high ranking.
2/.There is a serious amount of Page Rank manipulation going on, with links from high PR sites being traded for money.
3/.The idea that a link is a 'vote' on the quality of the site is no longer valid. To link to a site you need to know it exists. Most likely you find it on a search engine, or are advised of its existence by the site owner trying to acquire inbound links. In other words, it gives an advantage to sites that already score well in Google.
Of course, it is still better than the alternatives (with the possible exception of Teoma's variation on PR) - but it will need to change before it could claim to be truly objective.
Still a Google fan though,
4eyes
Pagerank is something that web site owners and administrators need to treat with respect.
Exactly right - but it shouldn't be, and that is why PR is flawed.
Site owners should not have to consider page rank at all - they should only need to consider the quality of their site. However, as more and more site owners are getting on the PR train and going looking for link exhanges, the whole system loses objectivity.
[edited by: 4eyes at 8:24 pm (utc) on July 27, 2002]
Ironically, had they kept the whole thing close to vest (i.e., SECRET) in the first place - hard to to when you're in University and publishing! - the PR idea would have had a long, long life based on relevant results.
"However, as more and more site owners are getting on the PR train and going looking for link exhanges, the whole system loses objectivity."
As the above quote is so, so true, its longevity will now be based on Google's willingness to change. I love it when people say that the Page Rank factor is becoming less and less important. What does that mean? Are we regressing to earlier SE strategies? I haven't heard of any conceptual breakthrough's on the hey-let's-index-the-world front. :-)
"However, as more and more site owners are getting on the PR train and going looking for link exhanges, the whole system loses objectivity."
As the above quote is so, so true, its longevity will now be based on Google's willingness to change. I love it when people say that the Page Rank factor is becoming less and less important. What does that mean? Are we regressing to earlier SE strategies? I haven't heard of any conceptual breakthrough's on the hey-let's-index-the-world front. :-)
Google could save the excellent Page Rank concept by simply removing the option to view PR from the toolbar. In retrospect I think the biggest mistake they have done was to make PR available on the toolbar.
I agree that it takes time for a new site to obtain a good PR level, but this said it also takes a new site time to develope it's content. After all it is content that generates links.
That said it is still possible for a new site to rank well in the serps. Pagerank is only part of the equasion. If a site is relevant it will be shown.
But I would like to discuss in this thread the PR algo 'as such'. I really think it will have to change dramatically during the 3 years to come. It looks to me like PR based on number of links and PR from those links is going to a dead-end soon. Adding 'themed' PR is one possibility, but extremely complex. Certainly when you consider that you have to have an 'intelligent' algo able to understand 50 languages? What could possibly be another, not exclusive, addition in the algo? Traffic? Time spent on a site? Voting values by visitors related to their search?
What would you consider to be good variables to include in the PR algo, apart from the link factor? And how would it work practically?
Thanks!
On average, an un-reciprocated link has more value than a reciprocated one.
Of course, if it were a simple filter, it could lead to a series of link 'rings' - so probably some kind of analysis that takes into account links from unreciprocated neighborhoods.
hmmm.... could be happening already guys;)
