Forum Moderators: open
In principle one can think of a number of modifications which could be implemented, i.e. the transfered PR could depend on:
- the position of the link / the total number of links
- text or image link
- the size and style in case of text links
- the size in case of image links
- internal or external links
- other parameters
I see clear evidence that a least small text links at the bottom of a page are treated differently.
Did anyone else found provable examples of modifications made by Google. (please, no guesses). Or did anyone try to calculate the relation of transferred PR for different kind of links?
As you suggest, various parameters about links are probably in the algorithm, link text being especially important.
Maybe we can continue your discussion by exploring your assertion that small text links at the bottom of the page are treated differently. Can you explain in more detail about what you've noticed?
I'm also curious about this, about whether you're suggesting that "size matters", "location matters", or whether it might be something like Google's ability to recognize "templates" of recurring elements on many pages within a site.
I observed that the transferred PR for small text links is less than for other text links, e.g. the "BestBBS" link on this site doesn't count as much as the "Home" link on this site. This is different compared to the original PR algorithm, where each link get the same weight (factor * PagePR / number_of_links). The reason could be that the weight descreases in the order as the links occur. (The total sum of transferred PR for all links will be probably the same, but the partitioning is unequal). Also the weight of a link could depend on the font size or other things...
(Of course, PR is just one component in the ranking system. Here I will just focusing on the PR algorithm, not on the ranking algorithm.)
Hopefully, my descriptions are understandable
I observed that the transferred PR for small text links is less than for other text links
doc-z, how did you zero in on this?
And, I'd like to actually see the proof with a real world example.
What happens if all the links on a given site are the same size and "small"? Would they all then count less than they did in the last index?
It makes more sense that links at the end of a document would be given a lower weighting and links at the top of the document a higher weighting.
But this is all speculation, isn't it...?
Has Google issued any sort of document yet on the modifications?
I have some page which should have the same PR since they are linked in exactly the same way. However, this is not the case.
ronin
of course, the theory that links at the end of a document would give a lower weighting than those at the top of the document is more likely. I just want to show the different possibilities to explain the behaviour.
To verify my observation, I built some pages with a special link structure (no content). The distribution of the PR will not only answer some of the questions, but also show the relation between different kind of links. Unfortunately, these pages don't have a PR so far. (I have to wait for the next update.)
One thing I discovered why using it, while examining my own site, is the strong importance of # of links on each page. Where I have for example, 5 pages in an obscure subdirectory, each with PR=5 on the toolbar, each linked sequentially, and each with reciprocal links to a a subindex page. I would have guessed all these 5 pages would have the same PR, and be ordered from 1 to 5 in terms of PR. But when I checked with the -link tool, I realized that there were differences in their PR order, which were explainable by a variation in number of outgoing links on each page.
Anyway, good luck with the experiment. keep us informed.
thank you for mentioning the "-link" syntax. I agree that the pages seem to be ordered by PR (not transferred PR). So far I found only few violations of this theory which could be caused by ToolbarPR problems.
If you have 5 pages which have the same incoming links and each of them linking to the other 4 pages, then these 5 pages should have the
same PR (independent from the number of outgoing links), if the original PR algorithm is used. Outgoing links decrease the PR of these pages (as well as the other ones), but the relation should be the same. Therefore, either Google has changed the algorithm or the 5 pages have all the same PR and are orderd by some other parameter (or appear randomly).
