Why Do Google Hit Numbers Vary?

Supa-Fly writes "I have a question about some conflicting results with the search engine google. I did a search for "pictures of mountains" and got exactly 1 million results. My friend did the same search (from the same office)and got 1,010,000 results. A second friend did the same search as the last 2 and got 1,020,000. These have not changed and every person gets the same results each time. My question is what is up with the discrepancies on google's search results?" Since this question is hard to answer from the outside, Craig Silverstein of Google kindly supplies his best answer to this question, below.

Craig writes: "Thanks for the great question. We get this from time to time and hopefully I can clear up some of the confusion. The number of estimated pages listed to the top right of a Google search results page is indeed, an estimate. It's a good estimate but still, an estimate.

There are many reasons why one might see a difference in the estimated number of pages returned for the same query. It's most likely the queries made by your co-workers were sent to different Google datacenters in what appears to have been a round-robin fashion. The index at any given Google datacenter can change slightly over the course of a day (each index is refreshed completely every three to four weeks). Depending on which datacenter finishes a query, the estimated number of results may vary.

Without having direct access to your environment it is hard for me to tell for sure, however, I believe this is the case."

wow, could we all have missed this?

[Tiber]

About a month ago, someone posted this story [kuro5hin.org] over on K5 [kuro5hin.org] regarding the google dance [internet-a...manual.com]. Good to see it's run by a marketing site, I couldn't think of anyone who might have more of an interest in rankings then those bastards. :P

Re:number oddities

[pete_p]

That's because Google doesn't do boolean searches. It will ignore the or (too common a word) and ends up treating it like an and search.

Re:Its too bad..

[daveinthesky]

We would all benefit from knowing which are the most popular searches. This, [google.com] though, is all google offers if I remember correctly.

Re:number oddities

[sparkz]

Wrong. OR is a boolean operator to Google. Check the "Advanced Search" link.

Re:Its too bad..

[danimal]

Ah, but Google does have one....well, available at the Googleplex [aaronsw.com].

Re:number oddities

[millette]

Actually, if you use an uppercase OR, it will perform a boolean search. Otherwise, the search defaults to an AND, unless of course you're using doublequotes "like this" to search for a phrase.

Re:Its too bad..

[Phroggy]

It's too bad Google doesn't have one of those things where you can watch everyone's search scrolling down the screen live. I bet there would be a lot of "pictures of mountains" searches right about now.

Try the Google Viewer [google.com] maybe?

Re:Its too bad..

[Tsuzuki]

From a vague memory of the last major /. article about Google, don't they have a censored realtime display of searches scrolling behind their reception desk?

And what you're thinking of is Metaspy [metaspy.com] - it still has that uncensored option. ;D

Re:Also cookies too

[feepness]

Remember cookies are kept in all browsers.

When a search engine finds those relating to their advertisers or 'favored sites' those are 'extended' into a higher level.

Some results may be discarded if certain advertisers don't like you to see those competing sites.

Uhhh, no. A domain can only see their OWN cookies.

Re:Its too bad..

[mrm677]

It's too bad Google doesn't have one of those things where you can watch everyone's search scrolling down the screen live.

No, but you can have Google scroll the results of your search!! Try this, and other "Google Experiments" at http://labs.google.com [google.com]

Re:Its too bad..

[oobar]

This page [searchenginewatch.com] at Search Engine Watch has a half dozen or so real-time search query viewers. It also has some quasi-realtime "search popularity" stuff too.

Re:Interesting Google phenomena

[dknj]

kuro5hin.org article [kuro5hin.org]
Google Dance [internet-a...manual.com]

-dk

Google does reverse-routing

[SHEENmaster]

so you may connect to any one of several servers. The servers each have different databases to pool results from and different caches to display.

A google search for my site [google.com] returns our old site that has had dead dns records for nearly a month above my new site [frob.us]. Sometimes my new site pulls into the lead, sometimes it isn't there, and at least one cache has the announcement that the old name was lost and a domain was purchased.

Google Dance

[kiwirob]

Results can also vary due to the Google Dance.

Google has 7 data centers each with a copy of it's index and these are "usually" mapped to www.google.com [google.com]. But google also has versions located at www2.google.com [google.com] and www3.google.com [google.com].

During the monthly update there can be different version of the index on each of the 3 versions. A website www.google-dance-tool.1hut.com [1hut.com] provides results for a search done on all 3 of googles index.

To check to see if the google dance is happening the most common technique is to check the "back links" for mayor sites like Yahoo by typing "link:www.yahoo.com" into the search box. this will list all the sites with links to "www.yahoo.com".

The Google Dance Tool site mentioned checks google every 5 minutes to see if the dance is on. Once it is started it sends out an automated email to subscribers (like me) so I can visit the site and see what the search positions for the next month on google will be using their google dance tool search.

Re:uh... (sex)

[Anonymous Coward]

A friend of mine has the 250th or so (it changes daily) hit for the word 'sex', and he regularly gets 100-200 hits a day from Google and Yahoo (combined). Sometimes people (horny people) go past the first couple of pages ;-).

More likely, those 100-200 people do a search on several words. Your friend should check their server logs to find out what people are really searching for.

Re:i'm not so lucky like you

[owillis]

Google uses Dmoz [dmoz.org] as a pagerank source. Just submit your site through Google's forms, I've had stuff listed within 1-2 days.

Re:googledance

[kiwirob]

Google Dance Tool is also available a www.google-dance-tool.1hut.com [1hut.com]

This site checks every 5 minutes for the google dance and has a alert mailing list for subscribers to get notified as soon as the google dance starts.

Re:uh... (sex)

[appleprophet]

And I have the 80,000 hit for the word 'sex' yet I still get 100-200 hits per day from Google and Yahoo.

I think people tend to do more complex searches.

Re:Google Dance

[Edmund Blackadder]

Try also the internet archive.

www.archive.org. It is pretty amazing but they do have pages that i remember from like 7 yrs ago.

Google tricks - was Re:uh... (sex)

[jetmarc]

a) did he very (by examination of the "referrer" tag in his logs) that people really went there from a search for "sex" (and not "sex" combined with other words)?

b) people might be tired of "google optimized" webpages and manually insert a "&start=250" parameter in the address line, to skip the commercial sites and browse to the less commercial ones.

Btw, I use the b) technique in Google-Groups to find old postings. The sort-by-date option only sorts from newest to oldest, and by modifying the page number you can directly go to the last page - effectively reversing the sort order.

Re:Some different results

[ahogue]

You may not have noticed, but the following probably appeared at the top of your results page:

"always" (and any subsequent words) was ignored because we limit queries to 10 words.

Thus, your last two searches (and however many more you tried to do afterwards) look identical to Google.