Yesterday, I mentioned that searching for the word "Yahoo" on Google (without the quotes) returns less than 40 results. Last night, on SearchEngineWatch at http://forums.searchenginewatch.com/showthread.php?threadid=3385, Googleguy addressed this oddity by saying:
I asked someone about this..
I asked an engineer about this. It's an interaction in our code with how we do result crowding (result crowding is when you do a search like [dell] and we show two results from dell.com, with a "More results from www.dell.com" link under the second, indented result). Since one of the things "&filter=0" does is turn off result crowding, that's why you see all the results if you add that parameter.
The results are all available by adding "&filter=0" to the url, or you can add
to the query to get the ondomain/offdomain results.
From talking to the engineer about this, I'm pretty sure that it doesn't occur very often. If you want all the results, it's just a mouse click to get the results without result crowding (by clicking on the "repeat the search" hyperlink that adds "&filter=0"), or to get to the onsite results (by clicking on the "More results from yahoo.com" link below the second search result). Either mouse click should give plenty of results that way.
Ok, let's examine this more. First, he mentions Result Crowding - a new term to add to the SEO Lingo to indicate the filter that prevents more than two results from the same domain showing up in the SERPs (one indented below the other). &filter=0 turns off result crowding but apparently it also has other functions, since GG said that turning off result crowding is ONE of the things that it does.
In the latest update which began in mid-December, many of the sites which lost rankings return to their original rankings with the use of &filter=0. Not all do, so obviously there is more than one change in the algo. However, because 3 of my sites lost rankings, and all 3 return to their original positions when using &filter=0, I have been focusing on discovering more about that particular algo change.
My gut tells me that it is some form of duplication filter. None of my sites have duplicate content within the sites, and copyscape.com does not bring up any flagrant duplication from other sites. It did show, however, one site that duplicated my meta description tag, and several sites that used my exact title and description for linking purposes. Still, neither of those things *should* cause a dup filter to kick in, but if Google over-tweaked the dup filter, that may be the cause. Of course, I've tweaked my tags to see if that releases the filter, but have seen no results yet.
Obviously, Google has decided that this problem is minor and not one that they intend to do anything about, so it will be up to us to find a way around it. I intend to try to find a few searches that return a small set of results, and then analyze the differences when using &filter=0 (i.e. other than the normal result crowding differences). If I can find a pattern, it may lead to a better understanding of everything that &filter=0 is used for, which hopefully will lead to what happened in mid-December in the algo change. I don't really expect to be able to reverse-engineer this, but the only way to be sure is to try. Of course, I'll share anything I may find.