[LINK] RFI: Backtracking Links

[rene]

On Fri, 28 Jan 2011 21:00:45 +1100, Roger Clarke wrote:

> At 19:05 +1000 28/1/11, rene wrote:
>> Suggest (if you haven't already done so) that you try playing
>> around with the range of "SEO tools" provided by the new search
>> engine: http://blekko.com
>
> This gets really interesting.

Indeed :)

> The expectations would be that:
> (a)  Google would have the bigger catchment;  and hence (b)  Google
> would generate the larger link-count

Yes, as I'd assumed, but evidently wrong.
[...]
> Google gives 118, which reduces to 53:
[...]
> Blekko gives 820, which reduces to 642:
[...]
> Spasmodic sampling, biassed to the later entries on the list found no
> spurious entries in either case ...

I ran some tests and associated analysis today in relation to a sub-page on 
my site (I used a sub-page - which has existed since approx Jan 2007 - so 
that it would be less time consuming to do more than spasmodic sampling, 
than would be the case with links to my site's homepage).

Google gives 31 inbound links which reduces to 15.
Blekko gives 157; or 98 or 96, inbound links depending on what, exactly, 
one 'asks' blekko.

 From a fairly detailed analysis I conclude Google's results are near 
useless, and certainly inaccurate in terms of how many pages actually link 
to the specified URL. I'm unable to determine any logical explanation for 
why Google's results are substantially less, and that's after taking into 
account questions/issues such as whether one or the other search engine is 
incl/excluding links depending on whether they're in page body text, or in 
follow up comments on e.g. blog or news sites, or in side bar menu links to 
so-called "blog rolls" etc.

Google results turn up 4 (of 15) that blekko does not show. Blekko shows 
over 100 that Google does not show. 

Blekko's results vary depending on what one asks, i.e.:

1. This:
http://blekko.com/ws/http://[page url] /links
shows 98 links which reduces to 96 when one clicks on the 'last' page of 
the search results listing.

2. This:
http://blekko.com/ws/http://[page url] /inbound?strict=1
shows 157 inbound links.

The reason for the difference between blekko (1) and (2) is by no means 
clear, but does appear to have some *loose* connection with where the link 
is on the found pages, *among other things*, and both show some pages [not 
always the same between (1) and (2)] that are basically copies of other 
pages eg: such as /blog/date/name; /blog/name; blog/archive/date.

Overall, I reckon (2) seems to provide a better list of arguably 'quality' 
pages that have links to the asked-about page, but I expect the (1) and (2) 
results are likely to vary depending on what page is asked about and what 
pages link to it.

The one thing that is abundantly clear to me is that some 5 or so years ago 
using Google's advanced 'links to' option did seem to produce a probably 
reasonably accurate set of pages. Some 1-2 years ago, it became apparent to 
me that it no longer did - because I knew more about what pages linked to 
some of my pages than Google was capable of showing. That's still the case, 
and now blekko does produce a list that shows not only what I already know 
about, but also quite a lot more - and that on sampling, do in fact contain 
a link to the relevant page.

Irene