Search Engine Watch
SEO News

Go Back   Search Engine Watch Forums > Search Engines & Directories > Google > Google Web Search
FAQ Members List Calendar Forum Search Today's Posts Mark Forums Read

Closed Thread
 
Thread Tools
Old 07-07-2004   #1
cariboo
Member
 
Join Date: Jun 2004
Location: Paris, France
Posts: 33
cariboo is on a distinguished road
Does Google still use pagerank algorithm ?

I don't think so...

I firmly believe they use blockrank now instead.

And I think they have been using it for months... All the changes seen in the SERPS after Florida update can easily be explained by the used of biased pageranks.

We know it is very difficult to achieve continuous updating of the Google index with the old pagerank algorithm... PR computation was very slow... But we can observe actual rolling updates know, and it means they have changed their algorithm. The best candidate is blockrank, combined with other techniques, like "the adaptive pagerank" all developed by the same team from Stanford University (the Kaltix team, enlisted now in Google payroll).

Blockrank can make pr calculations very fast, and it is also a good tool to build "topic sensitive pageranks"...

We can see two different kind of use for "topic sensitive pagerank" in Google Labs : "Site Flavored Search" and "Personalized Web Search". It means there are many thematic pageranks stored in Google index by now.

It is also very easy to use biased pageranks to improve relevancy on special queries... So why Google would not use this technique, which is at hand for them ?

I was puzzled when I read Matt Cutt declarations at the SES conference. Why receiving link from an authority site on a subject is important ? Why should we make links with other sites on the same subject ? If Google still uses the good old pagerank, then each link transmits its pagerank the same way, and the content of the page has no importance... But if they don't use the old algorithm anymore, then it is interesting to discover what kind of algorithm(s) Google uses now, or plan to use tomorrow.

I would be glad to hear your opinion about my theories and thoughts...
cariboo is offline  
Old 07-07-2004   #2
Mel
Just the facts ma'm
 
Join Date: Jun 2004
Location: Malaysia
Posts: 793
Mel is just really niceMel is just really niceMel is just really niceMel is just really nice
I am a bit confused, you asked about PageRank but appear to be talking about Googles ranking algorithm which are two different subjects.

As far as I know Google still follows the original idea of computing relevance first and then adding an increment to the rankings for PageRank which results in the final SERP listings.

The best evidence of this stiil being the case is that Googles PR and backlinks are updated about every three weeks while other aspects of Googles indexing and ranking seem to be almost continuous.

IMO Google maintains a public and a private (for lack of better terms) linking database, and excerpts the linking data at a particular point in time, uses that data to compute PR, with the new PR and backlinks then being set for public view, while in the background new data is being added all the time.

There is much more to Googles relevancy ranking algorithm than PageRank.
__________________
Mel Nelson
Expert SEO Dont settle for average SEO
Singapore Search Engine Optimization and web design
Mel is offline  
Old 07-07-2004   #3
bwelford
Internet Marketing Consultant
 
Join Date: Jun 2004
Location: Montreal, Quebec, Canada
Posts: 100
bwelford is on a distinguished road
I'm with you, Mel.

The private database is "real-time" and works really well. That's why Google serves up relevant and up-to-the minute information.

The public database that fuels the Toolbar and the link: search is synchronized with the private one on a very irregular basis and is very ropy. I always give the benefit of the doubt to a sharp outfit like Google that this is done intentionally (to fool SEO's) rather than out of incompetence.
__________________
Barry Welford
SMM Internet Marketing Consultants
bwelford is offline  
Old 07-07-2004   #4
rustybrick
 
rustybrick's Avatar
 
Join Date: Jun 2004
Location: New York, USA
Posts: 2,810
rustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud of
I think a lot of people are confused about what PageRank really means and does. I was speaking to one of my employees about PageRank and for some reason, he felt that the more relevant your site is the higher the PageRank. One really has nothing to do with the next.

PageRank is simply a vote from page A to page B. The more votes a page has (plus the more votes the pages that vote for you have) the higher the PageRank. So if you have a link from a page on flowers but your page is on computers, the PageRank vote is worth the same as if the link being from a computers page.

Google does not look exclusively at PageRank, if they did that would be a disaster. PageRank is one small part of the overall ranking algorithm.
rustybrick is offline  
Old 07-07-2004   #5
cariboo
Member
 
Join Date: Jun 2004
Location: Paris, France
Posts: 33
cariboo is on a distinguished road
Quote:
I am a bit confused, you asked about PageRank but appear to be talking about Googles ranking algorithm which are two different subjects.
No, Mel, I'm talking only about the algorithm Google uses to compute pageranks. What I say is that they don't use the old pagerank algorithm, but an improved algo developed by Kaltix : blockrank.

Quote:
The best evidence of this stiil being the case is that Googles PR and backlinks are updated about every three weeks while other aspects of Googles indexing and ranking seem to be almost continuous.
It doesn't prove anything. In fact, this topic has been discussed many times : backlinks and PRs are obtained from a separate database. What we call now a "googledance" is only a small update of the data used by the googlebar tool. The discrete changes you see in this special database are hiding the rolling updates of PR and backlinks for the entire index.

Quote:
The public database that fuels the Toolbar and the link: search is synchronized with the private one on a very irregular basis and is very ropy. I always give the benefit of the doubt to a sharp outfit like Google that this is done intentionally (to fool SEO's) rather than out of incompetence.
I agree with you bwelford ...

Last edited by cariboo : 07-07-2004 at 09:50 AM.
cariboo is offline  
Old 07-07-2004   #6
Mel
Just the facts ma'm
 
Join Date: Jun 2004
Location: Malaysia
Posts: 793
Mel is just really niceMel is just really niceMel is just really niceMel is just really nice
Hi Cariboo

IMO PageRank is such a small portion of the Google ranking algo that changes in PR are almost impossible to see in the SERPs, and for this reason cannot agree that the changes in the SERPs can be attributed to a different PageRank algo.

Google has said many times that they use more than 100 factors in computing the ranking of a page and PR is only one of them, and a very small one at that. As an example, after a previous PR update my PR and backlinks went down, but my rankings went up. Another example, one of my PR4 pages ranks higher than a Google PR9 page.

As I see it (others may disagree) in the beginning PageRank was an important part of the ranking algo, but it quickly became gamed by webmasters worldwide, forcing Google into finding newer ways of ranking the relevancy of a page.

I do agree that Google is interested in improving the pagerank portion of the algo, hence the patents for local rank, topic specific page rank, etc, but I see no evidence that these have in fact been been implemented, however if you have some evidence that they have been I am sure it would make a fascinating discussion.
__________________
Mel Nelson
Expert SEO Dont settle for average SEO
Singapore Search Engine Optimization and web design
Mel is offline  
Old 07-07-2004   #7
cariboo
Member
 
Join Date: Jun 2004
Location: Paris, France
Posts: 33
cariboo is on a distinguished road
Quote:
IMO PageRank is such a small portion of the Google ranking algo that changes in PR are almost impossible to see in the SERPs, and for this reason cannot agree that the changes in the SERPs can be attributed to a different PageRank algo.
You're perfectly right about this Mel. It's very difficult to see the effect of a single change in criteria use to rank pages in SERPS.

I talked about the "biased pagerank hypothesis" to explain the massive change in SERPS for competitive queries after Florida update. It is just another hypothesis, but it works as well as "Hilltop" or "a bayesian filter". If google implemented block rank to achieve continuous updates, it happened at the end of summer 2003 (we saw the first proof of a changed rate of updates at this moment). So they were able to use a biased pagerank to make an antispam filter for Florida update in november.

Why am I so sure about their use of Blockrank ? Well, because this algo is very fast (combined with adaptive pagagerank and other tricks, all found by Kaltix, it computes pageranks five times faster than the old algorithm). I haven't seen a clue of any other algorithm with the same characteristics... Other solutions exists, but only this one can be used without any drastic change of Google organisation.

Another clue : we found topic sensitive pagerank in use in the two google lab tools I've already mentioned. Well, it means you can find a set of topic sensititive pageranks in the Google index (i mean the dataset used to create SERPS in response to queries).

Well, but it's very difficult to compute biased pageranks : it is a slow process, and you must use a great computing power to achieve such a result. Unless you compute it on the limited transitional matrix of blockranks...
cariboo is offline  
Old 07-07-2004   #8
Mel
Just the facts ma'm
 
Join Date: Jun 2004
Location: Malaysia
Posts: 793
Mel is just really niceMel is just really niceMel is just really niceMel is just really nice
Yes Cariboo, but Google is still updating the public PR/links only once every three weeks, which it would not have to do if it were using blockrank, would it?
__________________
Mel Nelson
Expert SEO Dont settle for average SEO
Singapore Search Engine Optimization and web design
Mel is offline  
Old 07-08-2004   #9
donut
"No, no!" said the Queen. "Sentence first - verdict afterwards."
 
Join Date: Jul 2004
Posts: 48
donut is on a distinguished road
Terms still confused

Quote:
Originally Posted by rustybrick
I think a lot of people are confused about what PageRank really means and does.

PageRank is simply a vote from page A to page B. The more votes a page has (plus the more votes the pages that vote for you have) the higher the PageRank. So if you have a link from a page on flowers but your page is on computers, the PageRank vote is worth the same as if the link being from a computers page.
Pagerank isn't a vote. Pagerank is the measure of the quantity and weight of all the links pointing to that page.

Google considers a link from page a to page b to be a vote. So, Pagerank is not a vote, a link is a vote and PageRank is the cumulative measure of that vote.
donut is offline  
Old 07-08-2004   #10
cariboo
Member
 
Join Date: Jun 2004
Location: Paris, France
Posts: 33
cariboo is on a distinguished road
Quote:
Originally Posted by Mel
Yes Cariboo, but Google is still updating the public PR/links only once every three weeks, which it would not have to do if it were using blockrank, would it?
Oh, Google doesn't have to do it, whatever the algorithm may be

You can make a few interesting experiments to understand the way Google really works.

First of all, try to build a search engine, without a link based ranking algorithm, and using content based criteria... Well, on most queries, you obtain many pages with similar notes, and the "best rated" pages are not the more relevant pages... It doesn't work well.

Pagerank is only one of hundred criteria that Google uses to "rate" pages. But PR is a very useful criterion as far as relevancy is concerned. Because it measures link popularity, it prevents pages from being buried in the rankings just because they aren't optimized, and it prevents non relevant pages to appear in the first places.

The conclusion is clear : Google can't work well without any information about pageranks...

Well then, what about new pages ? And what about these very frequent partial (but extended) updates we have ben detecting for months now ?

These new or updated pages appear in the index, and they have their final ranking... So there are pageranks attached to these pages ! Now you have to think about it : what is the best way to compute a pagerank for new or updated pages ?

Another interesting observation to make : Google updates its index more often now. These updates are more frequent than PR and backlinks updates. But do rankings change when Google updates the PR or backlinks database ? Not at all !

That's odd ! PR does have an effect on relevancy and rankings. You change PR, and you can't see changes in the SERPS ...

The explanation is simple : the PR and backlink database is not the database in use to rank the index. It shows PRs computed days ago, and already in use for days. It serves only for two other purposes : the googlebar, and the "link:" functionality. And perhaps to put Googledance fans off the scent...

I noticed also Matt Cutt announced the end of "googledances" and, at least, that PR/backlinks updates will become more frequent in the future.

Anyway, these rolling updates are a puzzle : how google managed to achieve it ? Block rank is a good explanation. But I can take another one, if somebody has a clue.
cariboo is offline  
Old 07-08-2004   #11
Mel
Just the facts ma'm
 
Join Date: Jun 2004
Location: Malaysia
Posts: 793
Mel is just really niceMel is just really niceMel is just really niceMel is just really nice
I do not think we have to abandon all links if we want to do without PageRank, anchor text has been with us from the beginning of Google, and is a much more important relevancy factor than PR IMO.

Just use the same algo as now, but drop out the additional step of adding in Pagerank after the relevancy calculations are done and the SERPs might change a bit but they might not lose their relevancy much at all. Links would still play a big part, but it would not be "dumb" links like PR which is based on more the number of links from and to various pages without regard to the topic of the pages the links are from or the reason for the link. A link that from a PR5 page with only one outbound link counts the same in PageRank if the linker is providing it as an example of excellence or an example of what not to do.

I suspect that attempts to modify the famous PageRank may be more along the lines of correcting the defiencies that have been forced upon PR by zealous webmasters, with a solution such as local rank, which addresses the problem of internal links, multiple linked domains, and sitewide links in one go.

Quote:
Another interesting observation to make : Google updates its index more often now. These updates are more frequent than PR and backlinks updates. But do rankings change when Google updates the PR or backlinks database ? Not at all !

That's odd ! PR does have an effect on relevancy and rankings. You change PR, and you can't see changes in the SERPS ...

The explanation is simple : the PR and backlink database is not the database in use to rank the index. It shows PRs computed days ago, and already in use for days. It serves only for two other purposes : the googlebar, and the "link:" functionality. And perhaps to put Googledance fans off the scent...
Well as you will have noted I do not believe that PageRank has any affect on relevancy ( which may be why they calculate the relevancy first and then add on the PR) but the reason for not seeing a ranking change when the PR update rolls around is much simpler and is that the data used to calculate PR is IMO based on data at least two weeks old, with some or all of the time between when the data was extracted and the date it is published used for calculating PR.

The old Google dance updates seemed to be a shift from one set of stored page data to another, with possibly an algorithm change thrown in. Now Google seems to be able to update the stored page data incrementally, and the only "updates" we see are algo tweaks. All that remains is the more or less monthly PR update, and the reasons for that should be clear, it takes considerable time to do the PR calculations, and for that reason we are seeing an Update, which as Cariboo has noted, only confirms the suspicions we have had for a couple of week previously.

The other reason I don't think Google will drop PageRank is the same reason McDonalds will not drop thier golden arches, PageRank is synonymous with Google, and trying to rebrand that might not be possible.
__________________
Mel Nelson
Expert SEO Dont settle for average SEO
Singapore Search Engine Optimization and web design
Mel is offline  
Old 07-08-2004   #12
cariboo
Member
 
Join Date: Jun 2004
Location: Paris, France
Posts: 33
cariboo is on a distinguished road
Quote:
Well as you will have noted I do not believe that PageRank has any affect on relevancy ( which may be why they calculate the relevancy first and then add on the PR) but the reason for not seeing a ranking change when the PR update rolls around is much simpler and is that the data used to calculate PR is IMO based on data at least two weeks old, with some or all of the time between when the data was extracted and the date it is published used for calculating PR.
Well, I'm afraid you're wrong. PR has an effect on relevancy. But in combination with other criteria. Without a "link popularity" factor, any page can appear in the first place... Like in Altavista SERPs

If I follow you, it means Altavista had a chance to compete with Fast, Teoma, or Google, as far as relevancy is concerned. Well, good search engines rely on "link popularity" based algorithm. It's a fact !

Quote:
I suspect that attempts to modify the famous PageRank may be more along the lines of correcting the defiencies that have been forced upon PR by zealous webmasters, with a solution such as local rank, which addresses the problem of internal links, multiple linked domains, and sitewide links in one go.
Well, I know localrank are very popular among the US SEO community. But it is a very old solution (2001) and not an efficient one
It would be interesting to make public tests between the two algos (they have many features in common). I've heard of buddies who have done it. And block rank was the winner.

Block rank addresses well the problem of internal links, multiple linked domaines and sitewide links. Faster, better, and simplier. And almost without any human tuning. On the other side, with Hilltop or localrank, you should make many manual adjustments and add a great amount of external information to obtain good results.

Quote:
The other reason I don't think Google will drop PageRank is the same reason McDonalds will not drop thier golden arches, PageRank is synonymous with Google, and trying to rebrand that might not be possible.
I agree with you about that... So block rank or localrank can be good candidates, because it's still the old pagerank, used differently.
cariboo is offline  
Old 07-08-2004   #13
bwelford
Internet Marketing Consultant
 
Join Date: Jun 2004
Location: Montreal, Quebec, Canada
Posts: 100
bwelford is on a distinguished road
As a former statistician, I was always trained to stick with the simplest explanation until the data forced you to accept something more complicated. This approach was supposedly invented in the middle ages by a guy called Occam.

So here is my current view, whatever PageRank may have meant to those two Stanford guys in their bedroom office.

1. Search Engines pretend that the key factor in choosing one versus another should be relevancy.
2. Unless a Search Engine produces obviously irrelevant results, it is very difficult for anyone to do a comparison of different Search Engines on relevancy. So the key factor is unmeasurable.
3. Google has this PageRank factor that is "widely known" to be important in giving high relevancy results.
4. No other Search Engine has any such 'magic factor'.
5. PageRank is no longer used in the Google search algorithm (although complex information on backlinks is).
6. As a marketing strategy, Google maintains the fiction that their 'magic factor' is still the basis of their USP.
7. So Google does enough to keep an interest in PageRank as a concept. This explains why the Google Toolbar Pagerank is still published, even though it isn't kept up-to-date with any frequency.

Please let me know when the data comes in to force me to change my view.
__________________
Barry Welford
SMM Internet Marketing Consultants
bwelford is offline  
Old 07-08-2004   #14
cariboo
Member
 
Join Date: Jun 2004
Location: Paris, France
Posts: 33
cariboo is on a distinguished road
Quote:
4. No other Search Engine has any such 'magic factor'.
No other, but Fast/Alltheweb(a pagerank), Yahoo (webrank), Teoma, Webfountain(algorithms derived from HITS)...

Quote:
PageRank is no longer used in the Google search algorithm (although complex information on backlinks is).
How do know that ? I'm sorry, but it doesn't comply with your "Occam Razor" method.

Quote:
2. Unless a Search Engine produces obviously irrelevant results, it is very difficult for anyone to do a comparison of different Search Engines on relevancy. So the key factor is unmeasurable.
You're right when you say you can't make a direct measure of relevancy. But what you can measure is how people feel about the relevancy of a search engine... Relevancy is a subjective thing. But you can say "this search engine gives more relevant SERPs than this one.
You can't measure it, but you can rank it.
cariboo is offline  
Old 07-08-2004   #15
bwelford
Internet Marketing Consultant
 
Join Date: Jun 2004
Location: Montreal, Quebec, Canada
Posts: 100
bwelford is on a distinguished road
... I'm not sure the average searcher in the street can rank them. Or would even try. Everyone says you Google things now, so why look for something else.
__________________
Barry Welford
SMM Internet Marketing Consultants
bwelford is offline  
Old 07-08-2004   #16
Mel
Just the facts ma'm
 
Join Date: Jun 2004
Location: Malaysia
Posts: 793
Mel is just really niceMel is just really niceMel is just really niceMel is just really nice
Quote:
Well, I'm afraid you're wrong. PR has an effect on relevancy. But in combination with other criteria. Without a "link popularity" factor, any page can appear in the first place... Like in Altavista SERPs
I really believe you have too much faith in PageRank, which IMO really does nothing much in the Google algo. With PR any page can appear in the first place without respect to the PageRank of the page, or the content even the content of the page, Anchor Text back links can do that job all by itself.

The only place where I can even see PageRank as a serious factor in the rankings is in the event of two pages tieing in the relevacy rankings, in which case it could serve as a tiebreaker, and finding a case where the relevancy is identical and the PR different, would be like finding a needle in a haystack.

Search Engines are about relevancy, and PR has no relevancy factors involved in its calculation.

But it serves as the technical branding for Google (even though other engines use links in much the same way, and it also serves Google well in keeping much of the internet population fascinated by watching the little green bar, and it even drives up the cost of a link, so its a great public relations gimmick.
__________________
Mel Nelson
Expert SEO Dont settle for average SEO
Singapore Search Engine Optimization and web design
Mel is offline  
Old 07-15-2004   #17
Anthony Parsons
Rubbing the shine of the knobs who think they're better than everyone else...
 
Join Date: Jun 2004
Location: Melbourne Australia
Posts: 478
Anthony Parsons will become famous soon enough
You've taken the words out of my mouth Cariboo from discussions on this previously about PageRank and its current factors. I knew the algo was old, it was just how to word it. I feel better now, thanks mate.

So little time, so much to do....ouch. I am with ya Cariboo on this one. To many are stuck with older thinking caps, that things don't change. Nice work, keep it up.
Anthony Parsons is offline  
Old 07-15-2004   #18
Mel
Just the facts ma'm
 
Join Date: Jun 2004
Location: Malaysia
Posts: 793
Mel is just really niceMel is just really niceMel is just really niceMel is just really nice
If your thinking caps are set to include PageRank as a major or even important factor in Google SERPs then I would suspect that you may be strolling down the garden path, over the river and into the woods.

Any decent search engine is about relevancy and PR simply does not take any relevancy terms into account, which is doubtless one of the reasons that Google have always used other factors to calculate the relevancy of the page, and afterwards added PR on as a bonus. This is pretty clearly spelled out starting from the original design concept.

But Google continues to take the public stance that PR is the heart of their engine, and once having made that statement its hard to do away with PR without creating credibility problems, but its really easy to keep the term in the forefront of publicity releases, on the Google pages and staring every webmaster in the face from the google toolbar. But I am sure every one that frequents these forums is aware that a PR8 alone will not get you rankings.
__________________
Mel Nelson
Expert SEO Dont settle for average SEO
Singapore Search Engine Optimization and web design
Mel is offline  
Old 07-15-2004   #19
cariboo
Member
 
Join Date: Jun 2004
Location: Paris, France
Posts: 33
cariboo is on a distinguished road
Relevancy is not the point. We are just wondering what algo Google actually uses now.

Anyway, if you isolate only one criterion, no one is about relevancy. And there is no "relevancy factors" in any algos. But the combination of several (many) criteria, with subtile weight factors, make the ranking of the results possible, and you achieve a somewhat good relevancy with all of them.

Take anchor text for instance. You can have text in links with the same words as in the query, but linking to totally irrelevant pages... Well, if you use only one criterion, your search engine will give back very poor results.

As a matter of fact, a link popularity criterion acts as a "correction factor". This criterion is not more useful than others, and not less as well. But it's an interesting criterion to use in combination with others, because others are giving factual or statistical information about the content of a page, or anchor texts. Link popularity can give more weight to a page that is poorly optimized, but truly "relevant". And give less weight to a page that is optimized, but nobody wants to link to... The lack of correlation between link popularity and content analysis results makes pagerank a valuable correction factor.
cariboo is offline  
Old 07-15-2004   #20
Mel
Just the facts ma'm
 
Join Date: Jun 2004
Location: Malaysia
Posts: 793
Mel is just really niceMel is just really niceMel is just really niceMel is just really nice
Sorry cariboo, but search engines algorithms are about relevancy and search engines are about relevancy. Read the original Google design paper and you will note that google does all its relevacy calculations first and only then adds in PageRank. It not me thats isolating that factor but Google.

As a simple example of why relevancy matters and Pagerank does not, to even be considered for inclusion into the Google ranking pool for a particular word, you first have to have that word either on your page or in your links. That certainly is a relevancy factor IMO.

My points are simply that:
Relevancy is the altar at which all search engines worship, and most searchers too.
Pagerank is not about relevancy.
PageRank is IMO not an important factor in the present Google ranking system, as evidenced by the fact lower PR pages often rank higher than High PR pages.

If you are only looking at what Google is doing with regard to PageRank, then I suspect you are only looking at a very small portion of the ranking algorithm. It never crosses my mind when attempting rankings for a particular keyphrase to consider the PR of the page.

Linking these days is not the freely associated links that were the web when Google was created, linking these days is hundreds of companies set up to help you exchange links, software to find links for you, "advertisements" that will get you 50,000 links to your pages tomorrow, most sites boasting how many pages of reciprocal links they have secured etc etc.

The idea of counting these links as a way to correct relevancy I find either funny or sad, not sure which.
__________________
Mel Nelson
Expert SEO Dont settle for average SEO
Singapore Search Engine Optimization and web design
Mel is offline  
Closed Thread


Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off