Welcome to WebmasterWorld Guest from 188.8.131.52
Part of it seems to be the geo-location step for the IP address of the user. I've been wondering lately if the human editorial correction slips in at the last minute, too, but I just can't figure out how to test for that.
PS Just checked this on a DC watch tool and see that some DCs are currently showing different Allinanchor results to the majority, for the main term I watch. eg 184.108.40.206 220.127.116.11 18.104.22.168 22.214.171.124 all have different results to the majority
I truly believe that the top 10 are hand edited with assistance from the algo.
No doubt about that at all. The algo's "pure" results are then human reviewed and the results are tweaked for what the Google 2006 patent calls "favored sites" and "non-favored sites". Here's the discussion we had [webmasterworld.com] when that patent first came out, and the reality of human input into the rankings is very much alive today.
I'm just not sure whether the editorial factor kicks in when the data center rankings are calculated or afterwards.
1. At the beginning of all filtering, in order to provide some start juice that would be passed down in the iterative process similar to PR passing. (I think it is unlikelyI
2. At the very end when all algo-ing and filtering is done, this is very likely is happening as IT COULD BE APPLIED IN FAVOR OF LOCAL/COUNTRY CRITERIA, which seems to play a major factor in today's SERPs. Switzerland and Germany have so totally different results in the top ten, it astonishes me every day.
"For quite a while now, on some searches at least, there has been NO data center
that gives the same result set as google.com"
I see this too - and although I don't have any real extensive data to support
my gut feeling, it seems to me that I see the same results in DC's as I do
in Google.com for commodities, but for areas like news and services I see
different results.. and it's oddly capricious, too.
Just my $.02