This is false for other search terms. Search for "meaning of life". At the top left of the first result page it says "About 10,620,000,000 results (0.56 seconds)".
Screenshot: https://imgur.com/a/SQq7Snl
Logged into Google, Safari 15.6.1.
edit:
Here's a good one:
"starfish buttercup pickle mouse"
Page 1 of 2 says
> About 589,000 results (0.62 seconds)
Page 2 of 2 says
> Page 2 of about 22 results (0.36 seconds)
Oops! Must have been one of those common "multiply by 26,772" errors.
At the bottom of the page, it says "In order to show you the most relevant results, we have omitted some entries very similar to the 22 already displayed." with an option to show all results.
With all results shown, the total result count goes to 59, so only off by 9,983x! But who's counting? Definitely not Google.
What I get for: "starfish buttercup pickle mouse"
is exactly one result [About 1 results (0.32 seconds)], which is your YC post.
Removing the quotes gives me [About 34 results (0.63 seconds)].
The difference is that I have setup Google to show 100 results per page by default. I believe that this allows the (very approximate) results estimation algorithm to be more accurate. When I switch to 10 results, I get [About 591,000 results (0.53 seconds)]. Feel free to verify that this is the case for you as well.
There is no foul play here. I work at search serving, I've seen that algorithm. My personal opinion is that nobody cared to improve it because almost nobody (modulo YC crowd) cares.
In other words, the product is shitty because no-one at the company really cares about it - because of course, that “product” is just a vehicle for a big ad company. On top of that, there’s no meaningful competition. Hopefully we’ll get some DoJ action on this at some point.
I'd highly recommend this exercise for everyone who complains that there's no competition for Google, because a.) you're actually trying the competition and b.) you might realize how much Google does for you, and how far ahead they are against the competition. Who knows, you might even find a better search engine, and then next time you can come make recommendations for other HNers.
Most daily usage now I use Bing for the rewards points, and it’s approximately as good as current Google. Never have I failed to find something with Bing I then found with Google, despite it being a regular occurrence.
Yes, if only one page is shown, it’s accurate. I picked that (unquoted, sorry) phrase because it resulted in two results pages for me, showing a worse case scenario. With your settings, it would require a term with > 100 results to see the fiction.
I actually do assume it’s intentional, because
1. It’s predominant, at the top left of the page
2. I assume your UI team is competent and cares about every pixel shown on the screen, especially predominantly placed ones
3. it’s great marketing
4. It’s absurdly incorrect. 10x could be ubderstandable. It’s off by nearly 27,000x error here.
I can guarantee you that it's not intentional. It was definitely easier to estimate in the past, when a single query was fired on a single corpus for a given user input, but modern web search is vastly more complicated than this.
Both typed into the "url bar", and not search box on goolge.com. The only difference as far as I can tell is I copy/pasted purpose of life, and typed out meaning of life... macos & chrome.
What do you prefer for search if not?
But others have pointed out that those index numbers don't resolve rationally to results.
It seems the index number is a different match than what they choose to show.
I suspect (it was introduced before I started, and re-introduced after I left) that it's there to remind users of the work that software performs on their behalf, much like how TurboTax says "$X in taxes saved" or Brave says "9 hours saved blocking ads" or your brokerage says "Total profit: $X". Users tend to forget this, and just assume that they're entitled to the world where the product exists.
If the number was in the thousands in the late-90s, you could probably get through all of the sites.
If the caching was put in later, say, the early 2000s when google started getting really big, the number might just keep growing until it eventually looks ridiculous.
I've never worked at google, but I've seen a few systems that broke in weirder ways when nobody bothered to maintain them after they had outlived their usefulness.
Such a shame for basically one-trick pony who doesnt understand that staying relevant long term means that trick must be and remain a damn good proposition.
Its true I eventually give up and do ie duckduckgo and if that fails I try to refine searches even more (but this rarely works since I already start with it as default)
It's very ironic that a search engine, which is supposed to be helpful precisely for the things that are hard to find, seems to actively sabotage your efforts at such deep searching.
I have so many complaints about your product, it drives me wild.
How many people even look at organic results versus the paid-for ads that display at top?
It should be illegal for anyone to be able to purchase ads for another company's trademark. Apple, Google, and Amazon are all extorting companies by forcing them to buy ads to protect their own brand. (My own brand is being attacked by a competitor in this way, and it's ridiculous!)
The only reason anyone uses these systems at scale is that third parties were available in aggregate early on to provide content. You built your product off of our backs. And now that the power dynamic has shifted, we're cattle to soak for as much revenue as possible.
It also seems like the only reason Google is dominant is bad behavior. Paying for default search engine status. Being the default in all of their other unrelated platforms. Achieving browser monopoly.
I've recently started seeing Chrome ads and billboards everywhere. Google purchased a huge percentage of my city's billboard ad inventory for their "better on Chrome" campaign. It's as if Google knows this is the reason for everything. Where except for Apple devices is Chrome not dominant?
This whole cartel needs a muzzle.
Device companies should not be ad companies.
Ad companies should not be service companies.
Service companies should not be content and production companies, since they can favor their own and price pressure the rest.
We have a world where the top tech conglomerates are all of these and then some. They've cast a wide net and turned the whole world of consumer interaction into a supermarket, where we now have to pay for "shelf space" to interact with customers, pay to protect our brands from unfair sniping, pay to grow, obey asinine rules to build a product that fits their desired shape, integrate with their payments and login stack (so we're even less in a relationship with our customers).
It's a far cry from the open web of the 90's. Really bad for small companies, new startups, and even consumers. We can barely afford to build our products with all the margin that goes to Google, Apple, and the rest.
I wish the rank and file could feel this. :(
The way Apple and Google have wedged themselves in as an interface and then gatekeeper to our customers is deeply pernicious. I respect what they have achieved but it’s deeply saddening to wonder if the freedom we had decades ago online was anomalous and not intrinsic to the medium.
'Don't mind us destroying society, we are just following incentives!'
Swear I've heard that somewhere before though...
You are confusing users with queries. A lot more than 0.1% of users use the other result pages - double digit percent certainly. But if Google is doing a good job they don’t do it very often.
"Page 14 of about 184,000,000 results (0.83 seconds)" "Page 15 of about 149 results (0.76 seconds)"
"In order to show you the most relevant results, we have omitted some entries very similar to the 149 already displayed."
Maybe it's time Google spent a bit more time on their search engine, including pages 2-20.
Unlike today, right? Where actual search results... represent less than 10% of the page and the rest is irrelevant information and ads in Google search https://grumpy.website/post/0XCmMC-2O
In my opinion a search engine that can guess and understand my question from a single word is superior to encyclopaedic-bureaucratic return. (After all, in whole lot of times I am searching for something that I don't even understand what am I really looking for.)
> It'd make pagination basically useless if the same results you already went through show up on a later page because the ranking is different.
A bit off topic: I have always wondered why search result is the only feed on the web (apart from HN) that has not adopted infinite scrolling. This is a good explanation.
Extreme profiling Google? :-)
Search engines are still bad.
Having to go through 10 pages of results to find something that matches what I was looking for is one of the reasons I stopped using Google. What I want is almost never on the first few pages of search results.
And I'd bet the reason why it's still there (I left Search in 2014) is because < 0.1% of users ever hit the next page. Everybody else just refines their query to a different search. It's a holdover from when search engines were bad (i.e. around 1998) and you had to go through 10 pages of results to get the one you were looking for. As a result, Google expends approximately zero engineering effort on pages 2-20 of the results - I know that in the 4 visual redesigns I worked on, we didn't touch them once. It wouldn't surprise me if the response to flack on this is to just get rid of all pages other than the first one - it avoids the issue entirely and wouldn't affect 99.9% of users.
The technical reason for this behavior, as others have remarked below, is pagination. Ranking across the full result set is a very complex calculation, and it can depend on some factors that are basically random (eg. timeouts and failures in backend servers). It'd make pagination basically useless if the same results you already went through show up on a later page because the ranking is different. This requires that the full result set be cached. You can cache 400-1000 results for each of the queries that the 0.1% of users who actually hit "Next" care about, but you'd have a big issue caching 2 billion results for each of those queries.