[Web4lib] Google search results question

[Jeremy Dunck]

On 7/5/05, Felicia Mehl <felicia at u.washington.edu> wrote:
> I took a class on search engines where we did various searches on Google and
> compared results. We noticed, like you did, that the number of actual
> results was not the same as the estimate, also that each person got
> different numbers of total results although their searches were exactly the
> same (as per the assignment), and that using the "OR" operator seemed to
> generate a total number of results that didn't make sense mathematically.
> Our prof said that in some cases it was because these numbers were only
> estimates (based on how Google indexes documents), 

It's pretty common, when working with large datasets, to trade off
accuracy for performance.

Since almost no one looks past the first few pages, the total number
of results doesn't matter to most people.

I think the "about foo" bit comes from a statistical estimate based on
the number of hits within a smallish sample.  So if your query is
statistically unusual, your estimate will be off.

It does seem that they're maybe over-estimating, but I can't blame
them for estimating in general.