algorithm changes?

[pawel]

marinibuster,
what sort of algo changes did you write in [url] [webmasterworld.com...] [/url] about?
GG algorithms don't change too often,do they? In fact, the main algo remained unchanged ever since it was introduced in 1998 or so - it's just that coefficients in the PR equation could have changed, and human interference is ,naturally, various in different cases..

[doc_z]

I cannot answer the question for someone else, but you will find a comment about changes of the algorithm here [webmasterworld.com] (msg #12).

[dwilson]

Coefficients in the PR calculation can change, as you said. Also, coefficients in the rest of the algorithm can change. New elements can be added & old ones dropped. For example, several people have posted the opinion that alt text has either been dropped entirely or given much less weight.

Those are the kinds of algo changes we see from one update to the next. Which specific ones martinibuster had in mind, I'm not sure.

[pawel]

I just want to clear one thing out:
I was talking about PR calculation algo here, and only about this.
<final_page_ranking> := <page score> * PR,
where <page score> is calculated taking into account, as you wrote, alt tags, keyword density
and so on, and this may change over time.
For PR though, equation is pretty clear, and it measures *only* inbound links, weighted
according to those outer sites' PRs.
You can try your best, build a killer page and use all the tricks you come up with,
but if you don't have good linking to your site, your PR will be close to 0, and so your
'final' page rank - puting your page at the very bottom of SERP.

[Mohamed_E]

You can try your best, build a killer page and use all the tricks you come up with,
but if you don't have good linking to your site, your PR will be close to 0, and so your
'final' page rank - puting your page at the very bottom of SERP.

pawel,

At some level I suppose that I agree with you, BUT you seem to be taking an either/or approach. Yes, with zero links you are dead, and with far fewer links than the competition you will do badly. But with slightly fewer links than the competition you can do quite well by out-optimizing them.

In their basic paper on The Anatomy of a Large-Scale Hypertextual Web Search Engine [www7.scu.edu.au] Page and Brin write:

Combining all of this information into a rank is difficult. We designed our ranking function so that no particular factor can have too much influence.

We regularly get posts here complaining "Why does this PR5 site beat my PR7 site?" PR is important, but it is not everything.

[doc_z]

<final_page_ranking> := <page score> * PR

I think the final_page_ranking is a more complex function of 'page score' and PR. PR isn't as important as it looks in your formula. Also, you neglected the probably most important factor: anchor text (which shouldn't be part of 'page score' since it is an off-page factor).

[martinibuster]

Ok, that was a comment in a different thread.

I was talking about a manual tweak, in relation to Google specifically targetting one web site, in that case, Searchking.

The author of the thread thought he was under a black list, and I was commenting that that was highly unlikely.

[pawel]

doc_z,
you're right about anchor text.
"Anatomy.." [www7.scu.edu.au] gives a quite clear explanation of this - and I doubt if there's anyone else to know more about G algos than its creators.
Still, I claim that final_page_rank is computed pretty much close to the way I put it. Notice it embraces everything that's been said in this thread so far: PR5 can still beat a PR7 page, if only it uses good enough optimization techniques/tricks so as to offset lower PR factor.
What I tried to say is that PR is a multiplier in the final ranking - meaning, it doesn't *add* to page's rank, but rather this rank is eventually *multiplied* by it. To check this, simply do a query on GG and see PRs of the very bottom pages of SERP. They're PR0.

[doc_z]

pawel,

I didn't claim that I know more about Google's algorithms than their creators. (Although Google made a lot of changes in the ranking algorithm as well as in the PR calculation compared to the original papers.)

Please, don't post any well-known paper, exect you are one of the authors. I'm familar with these publications.

What I tried to say is that PR is a multiplier in the final ranking - meaning, it doesn't *add* to page's rank, but rather this rank is eventually *multiplied* by it. To check this, simply do a query on GG and see PRs of the very bottom pages of SERP. They're PR0.

You are probably right that PR not simply add to page score, but it doesn't seem to be simply multiplied. Also, you have to say if you are refering to the real PR or the logarithmic ToolbarPR.

[pawel]

doc_z,
I meant "real" PR.
Not citing well-known papers ;) "PR explained" tells quite a lot about the process of computing PR.