Impossible Counts

[randfish]

http://www.google.com/search?hl=en&l...2hard+money%22 - 645,000 results
http://www.google.com/search?hl=en&l...22+%2B+l oans - 846,000 results
http://www.google.com/search?hl=en&l...22+%2B+m oney - 745,000 results

How can there be more results for a phrase in quotes + another term than just for the phrase in quotes by itself?

[Dave Hawley]

There isn't, it's only Googles estimate that is wrong.

[Mel]

The problem is not with Google but with the formulation of your search. When you search Google for
"hard money" + loans

you are asking google to find pages that have either the quoted term OR the term loans on the page so naturally there are more results.

If you remove the space between the + and loans so that you are searching for
"hard money" +loans

then the search is asking Google to find pages which have "hard money" AND loans on the page and sure enough the number of results returned is less than that with "hard money" alone.

[dannysullivan]

you are asking google to find pages that have either the quoted term OR the term loans on the page so naturally there are more results.

No, that's not right. Even without the + symbol directly in front of a word, Google will still automatically try to find that word and ALL the others. By default, Google is an ALL search engine, not an OR one.

The oddity is real -- it shouldn't be that way, but I recall asking Google once what was up and got some answer I'll have to dig out on it. I'll see if I can find it and come back.

[Mel]

Right you are Danny MY Bad. - when you search for
"hard money" + loans

you are searching for

"hard money"
AND
+
AND
loans

which is made clear by looking at the search terms at the top right hand edge of the page where all three terms are shown.

[aixtal]

I've posted this about Googlean logic (or lack of) a few days ago:

http://aixtal.blogspot.com/2005/01/w...-logic-en.html

Counts are unusable.

Note that Yahoo! behaves in the expected fashion wrt Boolean operators.

[dannysullivan]

I've renamed this thread with a title of a new one aixtal started to make it clear what it's about -- may get some other people over to contribute. Aixtal, I also killed off that other thread as had the same info you posted above. Everyone, please carry on. I'll try to dig up that Google info I had to post later on.

[aixtal]

Thanks for cleaning up!

Actually, my second post was related, but the URL different. The one above is about the weird Googlean logic issue. The study behind the second link:

http://aixtal.blogspot.com/2005/01/w...nts-faked.html

shows that the counts themselves are screwed up, even if you don't use boolean operators.

It is likely, though, that the two problems come from the same cause(s), and it thus makes sense to merge the discussions.

[dannysullivan]

I threw a new blog post up, Questioning Google's Counts, that provides some more background on issues with search counts over time, if anyone wants history.

[randfish]

Was there ever resolution on this issue - why would Google's counts be so far off? Where do they get the counts they display from?

Looks like the original poster must have hit a hiccup, the current estimated returned results are much more in line for those three searches now.

http://www.google.com/search?hl=en&l...2hard+money%22 412,000

http://www.google.com/search?hl=en&l...22+%2B+l oans 485,000

http://www.google.com/search?hl=en&l...22+%2B+m oney 474,000


Side note: I have been seeing search queries like "keyword"+keyword2 coming up all over my log files for the last week or so, I was wondering about where they came from, now I know.