Actually I recently switched to DDG and the results are quite ok, especially programming/tech stuff. When not finding your stuff there is "g!". Fine for me so far.

This is exactly what I do and recommend. You use the private engine by default to see if it gets results. If not, switch over to one like Google. You loose the time it takes to do one search at worst.

Like you, I find it often gets me good results on programming or tech topics. Among other things.

While I understand your point, please.

> Anyway, we don't use DDG because its results are shitty.

These kind of comments are a bit unnecessary, especially on Hacker News, I believe.

Why? The topic is whether or not to use a search engine. This one delivers bad results compared to major ones. Frustratingly bad if you use it enough. I expected such a reply.

There is always a way to show that a search engine is bad with constructive feedback rather than just replying it's "shitty".

At least, the OP could have elaborated on why does he thinks this way so that everyone could understand why DDG's results are "shitty".

Unfortunately, no one can know what the OP thinks. As a result, it is hurting the work of others without any feedback so that they can improve themselves (i.e. DDG team) and serves little purpose (no explanations on the why) except expressing a frustration.

At least, that's my opinion. Let's be elegant in how do we relate to others. Feel free to disagree with me, but I think this kind of behavior is what creates a more toxic environment in our fields.

Alright. I even agree for most comments albeit less for this one. Let's try it your way so the feedback is specific, constructive, inclusive to audiences unfamiliar with grading search, and not profane. Here's my similar experience using the DuckDuckGo, search service after being a long-time user of the Google, search service.

1. I type search terms in text form into Google in an attempt to get a list of results that meet my needs. It virtually always gives me high-quality results I need in top few hits. High-quality means the results help me find what I'm looking for or better understand what my search terms describe. I do this into DuckDuckGo, it often provides low-quality results. Low-quality means the results were not relevant to the text I typed in, they were barely relevant to the text I typed in, or they were relevant but had less utility than results from Google.

2. I do exploratory search with Google by typing in words I think are related to the topic I'm trying to understand. It often gives me high-quality results in the top. If I do it with DuckDuckGo, it gives me low-quality results about as often as high-quality results. I see few patterns in the input that would let me accurately gauge whether inserting specific terms into DuckDuckGo instead of Google would produce as many relevant results I could utilize. This makes me unable to know if use of DuckDuckGo will have as high utility for the time and effort invested vs using Google's service. So, as I default on privacy, I had to try inputting the search terms into DuckDuckGo first with expectation it would often not meet my goals, mentally track what it succeed or failed on, and slowly gain instincts on what search terms to skip DuckDuckGo on in favor of Google to be more time-effective. As a Google user, I didn't have this problem that forced me to profile the search engine to gauge when it was capable of performing a search with high-quality results.

That, in detail, is why I had a poor, user experience with the search service provided by DuckDuckGo. Of course, explained constructively & in detail, it just looks like I think readers on Hacker News are idiots. Even my grandmother knows what a "shitty," "terrible," or "bad" search engine is. Such terms are how laypeople universally describe search functionality that fails to turn their queries into actionable information. They also short-hand what's common knowledge to save time. Describing your search experience in detail to such people will even sound condescending. So, instead I just usually say it's a a hit and miss search engine with result being them nodding their head in understanding. The other commenter apparently thought so, too, condensing it to one word and I still knew what it meant.

Thank you for the long response. I have done similar comparisons dozens of times, but only using search terms that duck duck go performs poorly on, and have had the opposite experience (this is a much harder test for ddg to pass than what you describe, making your comment more surprising).

What search terms are you using? Ideally, could you provide some searches that perform poorly on ddg, but well if you prepend g! to the term (which runs an anonymized google search)?

If the difference always comes down to google search personalization (so you can't provide a search others can repeat), that would also be interesting to know.

Here is the only query I ran in 2016 (so far) where ddg's results convinced me to use google, and then I found my answer on google:

drmSetMaster permission denied

The correct answer is to install xorg-server-legacy, and edit Xwrapper. I was looking for the now incorrect answer about setting some X binary setuid root. Both search engines provide lots of very out-of-date results, so I filtered ddg on setuid about half way through my search there.

Calling this query a "win" for google is questionable, but it is their top performer for me in the last year. Also the page with the answer does contain the string setuid.

Here's one last comparison for you that I just did while looking for FOSS tools for software testing. Difference is stark:

http://pastebin.com/j3v1NfYw

Ah ha! This is down to taste, I think. Google seems to rewrite 'FOSS tools for software testing' to 'Open source tools for software testing'.

Ddg does not. I strongly prefer the ddg behavior. To each their own.

"What search terms are you using?"

Don't even remember. It was mostly technology, history, academic papers, games, movies, and so on. I'm someone whose good at working around search engines but has a bad memory. I'm unfortunately not a good test case for the input as I get good at unconsciously working around problems to make a better baseline. I did have to regularly switch over to Google, though.

"but well if you prepend g! to the term (which runs an anonymized google search)?"

I didn't know about that feature at all. Thanks for telling me. I was going to run search through DDG and Google next time I have trouble to document the difference. Now, I'm considering doing DDG, DDG w/ Google option, and Google itself to see what the per-user customizations bring vs anonymized Google.

"If the difference always comes down to google search personalization"

I didn't think about that angle with Google. Surprised I overlooked it. I don't think it was affecting my search as a lot of it should have a large number of searchers. The esoteric stuff definitely got better results in Google over DDG's default, esp CompSci papers. Before I had Search Link Fix, I would type them into DDGo to get an actual, non-Google link to the PDF. This usually worked but some papers didn't exist in DDG at all. Even though they probably should've indexed the site (eg a university). So, it was clearly a deficiency. Other times, it was an obscure site that's been around a while but maybe low traffic dropped it out of a cache or something. Google had it since they store pages & have been for long time. (shrug)

So, maybe I can collect these into a little pile of files to drop on the team for their debugging.

"so I filtered ddg on setuid about half way through my search there."

Worst-case scenario. It's happened to me on occasion when a common term, esp reported in any news outlet, flooded my results with garbage. Eliminate it hoping to filter the noise. Usually works but this time the signal went out with the noise. I'm not so harsh on search engines with things like that since people in support forums drop all kinds of jargon that may or may not be related during discussion of a specific problem. I think that causes the bad results.

Out of curiosity, though, I ran your example. Your example is weird. They get the same top result with next best on one being a StackOverflow answer while another looks at StackExchange. I don't know what you were trying to do where I can't gauge the effectiveness of the answer. Here's a pastebin with the top 5 results I'm getting as of today:

http://pastebin.com/c9VGrT87

Did one or both have the answer?

There was lots of overlap of wrong answers. There were highly ranked correct solutions for older ubuntus from 1H 2016 and 2011-2013, I think. Definitely worst case for search engines.

The comment does not present any supporting evidence (even one query where ddg falls over would help, a deep technical analysis of why google is better, with large scale comparisons of result quality would be persuasive).

It is the equivalent to me replying to you with:

"You don't know why the comment sucks because you also suck."

Instead of pointing you to this wonderful Paul Graham article:

https://news.ycombinator.com/item?id=149052

> The comment does not present any supporting evidence (even one query where ddg falls over would help, a deep technical analysis of why google is better, with large scale comparisons of result quality would be persuasive).

While these are all true, we're on a message board, not debating in Ancient Greece. I am very sympathetic to calls to improve comment quality, but on the other hand it's unrealistic to expect someone to perform a deep technical analysis of why Google is better... in a message board comment.

Exactly. The default of the major search engines is we type in what we're looking for and usually get results. The commenter had a different experience with DuckDuckGo enough times to avoid it. We immediately know it gets inferior results to the status quo search engines. Many will avoid it on that basis. Other commenters might show up with detailed information on what kinds of queries succeed or fail. People wanting to use a mix of search engines, esp surveillance-oriented & private, might find that information helpful.

Yet, the original comment was sensible just because of two things:

1. Everyone on the Internet that uses a search engine knows what a good or bad one is. It has one purpose: turn your query into information you can use. If people say it doesn't or it's "shitty," then it's not doing it's job as a search engine effectively. All we really need to know.

2. The more specific information is better directed at the DuckDuckGo support or development teams to help them improve their alogrithms. Posting piles of data in Hacker News comments would only seem useful for the use-case I described above, people studying effectiveness of search engines via terms + results (better done elsewhere), or people reverse engineering DuckDuckGo's algorithm. I don't see why we'd expect people to produce or post that data as it's a lot of work and screen space.

1) Users do not know how to tell a good search engine from a bad one. There has been a lot of research on this, and it has been shown in many different ways. Here is one paper that is ranked highly by google scholar. It shows users cannot reliably tell a good result ordering from a bad one -- the result presentation is more important than the ordering. Other research has used more sophisticated techniques than "ask the user" to show that there are, indeed, "good" and "bad" orderings. http://onlinelibrary.wiley.com/doi/10.1002/asi.20941/full

2) There is at least one ddg developer here (not me), and HN is generally a good source of constructive feedback for startups. Maybe the original commenter's query string would help ddg or a competitor improve their product.

> Users do not know how to tell a good search engine from a bad one.

I am sure there is a presentation bias. I am sure that there are plenty of cases where a user cannot tell if a set of results are good or bad for a specific query. I am sure some users have trouble communicating why they think a specific result is good or bad.

However can people tell the general quality of a search engine? Absolutely. Google's simple aesthetics certainly helped them, but they didn't become the top search engine by luck and aesthetics alone. AltaVista, Yahoo, Lycos, Excite were all junk compared to Google, even the first version. It was very obvious.

I totally agree that google blew its competitors out of the water in the 90's. I switched to Google literally ten seconds after running my first query!

The research I'm citing is from the late 00's. At this point in the game, differences in search relevance are in the measurement noise.

1. Ill look at the scholarly results later out of curiosity. So far, the "found what Im looking for metric" is very strong anecdotally with widespread acceptance. Maybe Ill change my mind. We'll reject all opinions on the effectiveness of the engine in favor of whole, search histories on DDG and Google with terms and resulting HTML files as sole form of feedback in HN comments. Each user will do it for same terms for reproducibility or countering per-user generation biases. We'll discuss these specifics. This will lead, instead of claimed "good" or "bad" results, to claims of X terms or Y category having Z probability of accuracy or acceptance. Our discussion of DDG will be objective, scientifically accurate, possibly free QA for DDG, and likely be ignored by almost everyone everywhere (including most HN readers) outside CompSci or startups studying search engines. Im game, though, if there's a consensus here that we only discuss search constructively, objectively, and scholarly. ;)

2. DDG has contact info. I maintain that it's better to send data directly to them if one is troubleshooting it or helping them improve the product. Hoping a random employee is on same tech forum is a stretch despite it happening here more often than average. So, if we all did search data, we'd post it on a dedicated forum they host after we email them about the project.

Shitty based upon what? I would say they could improve, always room for improvement but DDG handles most of my searches with ease.

Why? It's a discussion about a search engine. The quality of the results are one of the most important aspects of it.

You're right that it's on-topic. But it's about as non-constructive as you can get while still being on-topic.

I disagree. Saying that the results are shitty is pretty on point.

>Well, no, they don't even have their own search engine.

This is false, as has already been pointed out.

If by 'shitty' you mean it has areas for improvement then yes I agree; as does your tact.

And after 10 years or so of trying, you'd expect the results to be far better, and more people to use it.

I think the search engine space is very ripe for a startup, but duckduckgo isn't it IMHO.

> after 10 years or so of trying, you'd expect the results to be far better

Would you? Most search engines' growth, including Google, was fueled by advertising revenue. Obviously that part of the equation isn't really an option for DDG.

I wonder if they would be better off as a non-profit, like Wikipedia. I would donate to them.

At its peak, Yahoo!'s results were quantifibly better than Google's for over a year, but no one noticed/cared.

Google search has been "good enough" for years, and so has the competition.

Multiple companies have burnt billions proving that search relevance will not convert users on its own.

If you want to crush Google, you need some compelling reason for people to switch (voice? privacy?), or you need to maintain good relevance until Google screws up badly enough. Maybe they will fire half the search team during an economic downturn in the 2020's, for example. At that point, you can use cash reserves to poach most of the top 10% of the remaining team.

The problem with long games like this is that you might screw up or lose focus before Google does. This happened at Yahoo.

Switching industry sectors, Microsoft had tons of negative press about 8, 8.1 and then 10's force upgrade/privacy debacle. With a mediocre or better hardware launch, Apple probably would have cleaned up. Instead, Apple's MacOS lines happened to falter in the same years as Windows did. Now, Windows devices are earning lukewarm to positive comparisons against Apple devices.

> Well, no, they don't even have their own search engine.

Really? I've never heard this, do you have anymore information about it?

It's sort of a meta search engine that utilizes other search indexes. Here's a post from the founder.

https://duck.co/forum/comment/27893

Edit, that's about 5 years old. The current page on Wikipedia provides a more recent explanation.

"DuckDuckGo's results are a compilation of "about 50" sources,[42] including Yahoo! Search BOSS; Wikipedia; Wolfram Alpha; Bing; its own Web crawler, the DuckDuckBot; and others."

https://en.wikipedia.org/wiki/DuckDuckGo#Overview

They actually say hundreds of sources now on https://duck.co/help/results/sources and specifically call out Bing, Yahoo, and Yandex for search results.

It is a meta search engine. Note that they never mention the share breakdown of these "sources". What percentage of queries DDG itself answer? Probably very few.

Meta-search is a type of search engine that goes way back. They avoid the cost of maintaining a full, search service by sending the query to all kinds of other search engines, running some analysis on their results, and presenting a condensed form to the user optionally with value adds. DuckDuckGo's value add is privacy. Here's some others to give you an idea for what they do:

http://www.searchenginepeople.com/blog/10-meta-search-engine...

Basically DDG is at the mercy of their sources, they can proxy the results for now to the actual search engines, but what happens if lets say Bing and Yandex requires them to pass some extra stuff or cut the query hose otherwise? Will DDG use its own index? Please.. They have no resources to build an actual half decent search engine.

I agree. Being effective and independent will cost money. There some potential for a company that charges monthly or annually to provide search services to other companies with privacy. There's lots of search, like marketing or competitive intelligence, where companies would prefer that the terms don't go through a surveillance company. Initial results might do as DuckDuckGo is doing now with them incrementally building their own database as funds came in.

Not sure how marketable it is, though. Google is pretty dominant. Most companies already running their stuff through it.

The results are extremely shitty indeed, especially for tech and programming related topics. I tried DDG for months and I found myself running every search twice, normally and once with !g appended. That's beyond shitty.

sure, the results are shitty, but you don't support big brother by using it... which is its primary appeal, afaik.

And I could just not use a computer, then I wouldn't be supporting big brother either....not supporting big brother is not an excuse to have shitty results, at least for most people.