What is the base value of the PR logarithm scale?

Forum Moderators: open

Message Too Old, No Replies

What is the base value of the PR logarithm scale?

Any guesses?

SkinnyJoe

2:49 am on Apr 10, 2003 (gmt 0)

What is the base value of the PR logarithm scale?

Apparently, this is a secret value, and maybe it changes with time or based on conditions.

Has anyone cracked the code? Or are there any good guesses out there?

Thanks.

Chef_Brian

2:54 am on Apr 10, 2003 (gmt 0)

This is my own take on pagerank take it for what it is worth. I see way too many webmasters trying to "crack" the pagerank code, instead I simple like to look at it this way.

* Google says pagerank is good ;-)

* I need more of the stuff

Really that is the end of my google algo crack ... understand that pagerank does play a very important role in which pages make it into the top 10 of the serps but pagerank alone will not cure all.

Instead of trying to "figure out" the special hidden secret of the google algo and pageranking codes would'nt webmasters be better served by working on securing more incoming links?

Brian

jcoronella

4:21 am on Apr 10, 2003 (gmt 0)

Others in the forum have told me "about 6". I personally have nothing to verify.

(sorry can't find the thread).

tedster

4:28 am on Apr 10, 2003 (gmt 0)

I've read a few estimates (they diverge pretty widely from base 4 to base 8, with 6 being the apparent consensus) and I've also also read that the distribution of PR is only "roughly" logarithmic -- not truly, mathematically so.

But the question I have is this: what good does it do us, even if we know for sure? If I could see an angle here that could help my sites, I'd probably dig into the question a bit more deeply.

mil2k

4:37 am on Apr 10, 2003 (gmt 0)

Yes previously the consensus was near 5 now it is somewhere near 6. No it has no direct use in seo but just helps to keep things in perspective. The base should change as the no of indexed pages by google increases. HTH.

figment88

4:43 am on Apr 10, 2003 (gmt 0)

That it is a log transformation is a guess in itself.

It could be a threshold system such as:
99th percentile=10
95th-98th percentile=9
87th-94th percentie=8
etc.

It could be based on varience such as:
> mean + 5 standard deviation = 10
> mean + 4 standard deviation = 9
> mean + 3 standard deviation = 8
etc.

It could be done with arbitrary cut-points, such as hey guys lets call everything over X, 10 and everything between W and X, 9.

Based on needed computational power and elegence a log transform is a pretty good guess, but it certainly doesn't have to be.

SkinnyJoe

4:57 am on Apr 10, 2003 (gmt 0)

There is a lot of PR obsession, huh?

I guess the fun comes in watching your score change, like a game.

If my PR0 site goes to a PR1, I will consider that a win.

If I go to PR2 (or stay at PR0), my degree of excitement (or diappointment) will depend in part on the answer to this question of logorithmic scale.

Base 6 would be pretty significant.

mat_bastian

5:26 am on Apr 10, 2003 (gmt 0)

This sounds so fascinating, I really wish i had a clue what you were talking about. I guess I'll just follow along and see if I can pick something up.

My teachers always told me mathematics would play an important role in my life... I laughed and said for what?

I wish I woulda listened in school and this thread is really driving that point home.

NotePad

5:31 am on Apr 10, 2003 (gmt 0)

mat, get the google tool bar, available here:

[toolbar.google.com...]

might clear up what all the hubub is about ;)

mil2k

5:54 am on Apr 10, 2003 (gmt 0)

mat if you think this is difficult then believe me you won't want to read all the PR discussion threads. They are all alpha , beta , gamma and other greek stuff! ;-)
(I hate the word eigen. Maths was the first subject i dropped in college's second year.)

[edited by: mil2k at 5:55 am (utc) on April 10, 2003]

Munster

5:55 am on Apr 10, 2003 (gmt 0)

Has anyone cosidered the 'Google guy with a dartboard' theory, I know thats how we used to work out our web stats at ********.com!

[edited by: Munster at 6:21 am (utc) on April 10, 2003]

NotePad

6:00 am on Apr 10, 2003 (gmt 0)

no URL's dude

oLeon

7:07 am on Apr 10, 2003 (gmt 0)

My teachers always told me mathematics would play an important role in my life... I laughed and said for what?
I wish I woulda listened in school and this thread is really driving that point home.

Oh yes, when I stuck in the SEO business, I had the same problem. There are that many issues about mathematic I thought sometimes I should go to school again...

Back to the topic:

Base 6 would be pretty significant.

I agree after reading here.

killroy

8:09 am on Apr 10, 2003 (gmt 0)

Does Google have some info regarding how many SITES it has indexed? Tha might be significant. Maybe the PR is limited as such it simply says, top10 websites=PR 10 top 100=pr9 and so on.

Meaning that for every website that goes up one drops. And might mean that high PR are strictly limited in number.

Just another theory to dirty the waters...

ciml

7:48 pm on Apr 10, 2003 (gmt 0)

figment88 and killroy, I a logarithmic scale as used in the Toolbar is likely to reflect the percentile approach very well. If you read Jakob Nielsen's famous "Zipf Curves and Website Popularity" sidebar from six years ago you'll see a parallel. Alternatively, if you think of PageRank as currency then Pareto's law (the 80/20 rule) fits nicely.

> what good does it do us, even if we know for sure

tedster, I've asked myself that question often over the last 12 months. In theory it can help us to "optimise" our site structures, but as Chef_Brian points out there are more effective ways to make use of one's time. On the other hand, a precise model of PageRank within a site is a fantastic scale for testing other weighting factors.

There's no doubting that the consensus is six, but as a loan voice I say higher, much higher.

madweb

8:04 pm on Apr 10, 2003 (gmt 0)

5.43656
7.3890461584

My two slightly-wild but slightly-educated guesses. I still have nightmares about A-Level* maths though, so perhaps don't take me seriously.

*UK academic qualification roughly equiavlent of the US "Advanced Placement" program

BigDave

8:21 pm on Apr 10, 2003 (gmt 0)

ciml:

There's no doubting that the consensus is six, but as a loan voice I say higher, much higher.

I'll not argue with you. The lone two months of experiments that I have tried so far seem to put it at *probably* greater than 7. Possibly much greater. The experiments are made much more difficult by two unknowns, the damping factor, and exactly how good of a PR5 the top level page is.

I will continue to use a base of 6 in my examples, just to keep people from arguing that particular point instead of having them pay attention to the point I am trying to make.

sun818

8:22 pm on Apr 10, 2003 (gmt 0)

Wasn't there a post that measured the pixel length of the PageRank in Google Directory? And the measure if broke it down actually went beyond PR10? (i.e. Google is PR11?) - I can't find the post now. :(

dkoller

8:56 pm on Apr 10, 2003 (gmt 0)

Personally I wish the toolbar went into tenths - or changed to a 0-100 scale. Showing only 10-11 scores for all of the billion of web pages makes the bar kinda frustrating. (Like imagine if a teacher rounded all scores to 100, 90, 80, 70, 60, etc. - You couldn't distinguish student performance well at all because there are far too many students clumped together, and the working set in this classroom example is very miniscule, nowhere near 3 billion!)

Your Internal pages, ranking in the directory among same pr sites, etc. can give you a hunch of a more exact PR, but it would be better to just have the bar say flat out, hey you are a 6.2. As of now one can only guess and/or waste a bunch of time trying to figure it out. Am I 6.2, or 6.8 hrmpph. BIG difference. Now... if I'm a 6.21 or 6.29 I could really care less.

If google is gonna let us see PR, give us some precision!

Mohamed_E

9:01 pm on Apr 10, 2003 (gmt 0)

> And the measure if broke it down actually went beyond PR10? (i.e. Google is PR11?)

Here is the Google has page rank 11 [webmasterworld.com] thread, the link to the table of pixels vs PR is somewhere in that thread.

vitaplease

6:36 am on Apr 11, 2003 (gmt 0)

Whatever the log factor may be (e.g. from 5 to 8), I think there may be some nice playing around of Google with dilution factors (i.e. dividing the Pagerank pass-through with the number of links on a page) depending on many factors.

GoogleGuy

7:02 am on Apr 11, 2003 (gmt 0)

42. :)

toprank4me

7:55 am on Apr 11, 2003 (gmt 0)

Entirely possible GoogleGuy! Had the venerable Mr.Page, in the company of the indomitable Mr.Beeblebrox, been sipping the Pangalactic Gargle Blaster while giving final touches to the PR algo! I wouldn't be surprised if it were the official drink of Google as well!
Heck! we could all do with one of those drinks ourselves ...to soothe our rattled nerves waiting for the update...

Just Guessing

8:44 am on Apr 11, 2003 (gmt 0)

4.2 :)

ciml

12:43 pm on Apr 11, 2003 (gmt 0)

Now GoogleGuy, 42's only the answer to life the universe and everything. The log base of your Toolbar is far more important. :-)

Dave, I agree that it's essential to know the top level page's PR more accurately. I spent many sleepless nights over it.

Are we talking about the log base as z in PRa = log_z(d . (z^PRb / numlinks)), or as in the number of links on a page that will case the page to give exactly one notch less of Toolbar PR? (d is often written as 1-d)

Either way, I vote somewhere between GoogleGuy's answer and Just Guessing's answer.