tedster - 5:16 pm on Jun 5, 2010 (gmt 0)
If 400 of 2000 links suddenly disappear, how do you 'diff' the GWT spreadsheets to identify what's missing?
This is something like de-duping any list, except when there are duplicates you want to eliminate both of them. Then you're left with a list of only those links that have vanished. I personally might do this in some other application than a spreadsheet - but that's because I'm a bit weak on spreadsheets syntax to use in this case.
I heard a rumor several months ago to expect a major Google shift in the backlink area somewhere around June - and from some of these reports, it sounds like that rumor may have been on the mark.
Google referral traffic is down, but only moderately, and still above May 17th levels. However, non SE referrals have dropped dramatically, almost from every website that is linking to my site.
This might point to a re-weighting of backlinks that target THOSE sites, which then creates a secondary cascade effect to your site.
It's almost as if some adaptive alg is testing my pages by exposing them on certain days, then shutting them down if they don't rank well enough or get clicks.
You may be on to something here. In the Webmaster Tools query term report, I am seeing an assortment of different PAGES being ranked for the same keyword - a term that historically had just one "winner" page on the site. This is in addition the the ranking position shuffle that get resported.
Having these other pages rank, even some of the time, has impacted the "winner" page's dependable performance for the site involved. It's not "strong AI" as in Skynet, but it surely does look like an adaptive algorithm. There seem to be waves of changes that are undulating through the SERPs, and it could take a lot of data, Stieltjes integrals, Fourier transforms and the Reimann zeta function to analyze it all ;)