Forum Moderators: open
My impression has been that it is important to have related links as Google values them more.
I'd sure like to know which is correct.
Thanks for any info.
Anne
I think it may be an issue of whether the themeing of links is a factor included in pagerank, or whether its a seperate factor, that still affects rankings, but doesnt affect PR
Point B: Most people here believe that at some time link relevance will be a factor.
I believe that there is considerable disagreement in this community about where we are along the journey from Point A to Point B. Also how "link relevancy" will be defined once we reach Point B. The recent patent shows one possible way of accounting for (not measuring) link relevance.
Remember awhile back when we looked at one of those 'mystery' pages that ranked high for one of your keyword phrases -- without anything on the page? I noticed then and have seen it a few times since that links from pages with keywords in the title and elsewhere on the page *appear* to give a boost in the SERPs.
But this might not be a matter of Page Rank, maybe just something else that's considered in the algorithm. Don't think we'll ever know for sure.
Jim
Relevancy is important in good design and therefore is probably considered in many of the other factors in google's algo.
There is no reason to move it into the PR calculation and muddy those waters when there are much more important areas to apply it.
But it is good design to stick to going after relevant links, so it is what I recommend.
[Edit - I'm agreeing with rfgdxm1, not Tartan. The standard pagerank calculation is simple in concept and mathematical, but Google couldn't determine relevancy with any more success than they can translate from English to German - i.e. somewhat, but not reliably.]
[edited by: jomaxx at 9:27 pm (utc) on Mar. 6, 2003]
anne,
I had made the assumption that while linktext doesn't affect PR, the simplest way to count link text (from an se's pov) is to include it as content of the page to which it links, since that is what it's describing..
I think the original paper discusses weighting of Page Title, Headings, Font-Sizes it terms of relevancy.. and if that's the case there must be some weighting of incoming link content, so that it fits in with everything else.. In which case, if I were google, I would alter the weighting of the link content according to the PR of the page from which it comes..
ie, a number of important (highPR) pages that link to you with the text "Blue Widgets" might make you a bigger authority on blue widgets than a similar page elsewhere who mentions blue widgets a lot in their content..
.. at no point though does this actually effect your PR as such, but as PR is only one of '100' (?) variables in producing the results that's not a problem..
All this is _conjecture_, but I'd _assume_ it's the kind of thing they'd do.. ;) I'd be very interested to know if people agree with this :)
The problem is that it becomes difficult to tell if they do it, since you're "number of important links" is going to effect your PR in it's own right...
There's much more than just PR going into the ranking. Sometimes it is my impression that we forget that there are many other elements to the algo than a simple ranking by PR. It has been discussed at length that Google's target is relevance. Anchor text may be nearly as important as PR. Well, maybe it isn't today, but maybe it will be tomorrow. The weighting of elelements in the algo seems to get tweaked every month. PR is a relatively simple calc (although not with billions of pages) but the ranking of relevance is certainly not so simple.
To do this would mean that for every link, it would need to be compare algorithmically both the target and linking page. To do this for every single link on the Internet seems computationally non-trivial. And as someone else pointed out, trying to do so would have the problem that it would not be able to spot relevant links that are. For example, a page which always calls it the "FBI" linking to a page that always calls it the "Federal Bureau of Investigation". Using PR as part of the algo has the advantage that it is cut and dry mathematics that cannot err. Adding something computationally intense that is this fuzzy to the algo seems dubious to me.
IMO a "vote" from a non-related site is usually just as valid as one from a site in the same niche.
Exactly. Google's "vote" analogy for PageRank is that the more valuable or "important" a site is, the more other sites will link to it. There's no reason to change that to "the more other related sites will link to it."
That doesn't mean that relevancy of links can't at some point be included in the ranking algorithms, though. But just like all of the other factors that weigh into ranking, it doesn't have to be included in PageRank -- and it almost certainly isn't now.
True that would be the case, but that would equally be the case for pagerank in general wouldnt it?. In any case it would only apply where comercial and non comercial were competing against each other - probably not the case for most searchs?
I am reminded every day of why NOT to exchange with competitors. Two of my clients are listed on competitors' websites, claiming they have better products/price than my clients. I look at the logs every day to notice BOTH companies sending at least 20 potential customers a day to our sites! (shhhhh I don't want them to take us off their list)
Steve
Most definitely the case. PageRank by its very nature favors non-commercial sites over commercial sites. This is a quirky side-effect of PR. There are a number of cases where non-commercial and commercial sites target the same keywords. The only solace to the commercial webmaster is that while the info sites will dominate over them, these sites really aren't competitors. Thus, if for "widgets" the top 8 sites are info sites, if you are selling widgets and come up #9, that's OK because none of the sites above you are stealing your business.
I think you've hit the nail on the head. I know I kinda missed the point of the question, but I agree that the content or the 'site' isn't often important, but the content of the 'link' may well be..
rfgdxm1, et al.. I completely agree that commercialy relevant sites seldom have a motive to cross-link, even to the extent that I'd assume this sort of cross linking is a contributing factor in determining spam (along with the sites being on the same server, same content etc ;) )
However.. surely a link from a non-commercial relevant resource site to a commercial site might be very much more of a vote, and so making that site more relevant to a particular search term?
Is it computationally difficult? Wasn't there a thread last week about a patent made for a post-processing system, whereby you would take the top 100 results, and re-calculate PR, amongst them (a second or so's processing?) .. I may have got completely the wrong end of the 'ole stick again.. wasn't the patent suggesting exactly what anne asked?
[edit]..
found it.. :)
[webmasterworld.com...]
[/edit]
PageRank by its very nature favors non-commercial sites over commercial sites.
This is why I'd think a comercial site with a lot of interesting related information would be a good strategy.
Anne
This is why I'd think a comercial site with a lot of interesting related information would be a good strategy.
Exactly!
If a commercial site is also an informational site, you might find yourself getting links from some of those big .edu and other non-commercial sites. Having great content is how you start collecing links like you would not believe.
You can also exchange links with those competitors that you are closely related to. If you sell stoves, you can have links to sites that sell pots. If you sell boats, exchange links with a local marina.
If it seems more difficult, just remember that it's also more difficult for your competitors! Any improvement that you get and they don't will put you that much further ahead of them.
