If anybody is curious for the distribution of used searches between ddg/!g, I came up with this rough estimate for firefox:

  sqlite3 ~/.mozilla/firefox/*default*/places.sqlite "select substr(p.url,9,14) as search_backend,count(1) from moz_places p where p.url like 'https://duckduckgo%' or p.url like 'https://%.google.%/search%' group by search_backend;"

I fall back to google every 11th query, not counting maps searches.

I got:

    duckduckgo.com|14129
    encrypted.goog|345
    images.google.|1
    ipv6.google.co|1
    mail.google.co|2
    play.google.co|10
    translate.goog|2
    www.google.co.|1106
    www.google.com|2084

Rather interesting. I'll compare with my personal PC when I get home.

I fall back every 21st query. That said, Google doesn't always find the answer in those cases, either.

duckduckgo.com|5521

www.google.com|6

So... I guess that means I don't really need google for search?

I don't think it's a bad thing either. It gives DDG a very clear signal so that they can improve: "you are not giving me the results I want for these queries". For most other search engines people just switch.

You can use !m instead of !gmaps.

I've found that "!w", which searches Wikipedia, is extremely useful.

One thing that consistently blows my mind is how readily an external search engine creates a better index, with more relevant recall, by scraping the published static pages of Wikipedia, such that it outperforms Wikipedia's own search,

Wikipedia should be able to search itself better than any external entity, but cannot. Wikipedia, in effect, is blind to its own data, and can deliver less insight into it's own content than multiple external Search entities.

If I use https://en.wikipedia.org/wiki/Special:Search for anything, I get weaker results, than if I use search engines to locate wikipedia articles according to the same input string.

There's something wrong with that.

Search isn't easy and I would fully expect a company that specializes in search to provide better search results than an entity that simply has search as a feature to their main product. Since Wikipedia provides all of their pages for indexing, unlike Facebook for example, a good search experience would be an MVP for any competent search company.

That's a good point. Put another way, were I to run a local copy of a wikipedia mirror, right here in my own bedroom, I'd be putting myself eye-to-eye with any wikipedia employee or volunteer. I'd have the same tools in my hands, as they do. I'd be one person, a PHP script hosted by an apache process, and the wikipedia data set.

But, contrast that to Facebook, and the concept becomes more interesting. Facebook's search utilities are similarly unsatisfying, despite the massive resources of the company. This is probably not a blind spot, but more about the level of quality provided, for free-tier, unprivileged user utilities. Advertisers probably gain better reach, but perhaps without knowing exactly who they reach. Facebook's search tool can't be used to discern the capabilities or qualities of one or more of the indices Facebook uses to negotiate the landscape of their data.

> There's something wrong with that.

I disagree, I think Wikipedia is akin to a collective blog where people just dump information to. Indexing is hard enough that it becomes a full-time job to do it well, and that takes time and money.

Search is so hard that external search is necessary for many websites to exist outside of a void. If we expect blogs to have their own searching functionality, we end up with a collection of disjoint webs, and that would basically kill the web as we know it.

This seems typical of many built-in searches. It's halfway how things are indexed but also how things are presented: one of the worst offenders is ruby's documentation, where it's extremely cumbersome to get the latest version of a given item[0]. And that's a trivial search query.

[0]: https://ruby-doc.com/search.html?q=Array

I don't see why this would be the case. Google has the advantage of seeing the content on pages with inbound links.

The same is true for every forum I know of, in fact a lot of them now have a google search box that gives site specific results

And !a for amazon and !yt for youtube and !dpkg for packages and !imdb...

Definitely my favorite feature of ddg.

The only thing that annoys me about DDG is that I have to use !wen for Wikipedia searches. !w will return localized search on Wikipedia, and I almost never want the Danish wikipedia.

I guess this is based on the Accept-Language header of your browser (which you can control), not guessed from your IP (which is a terrible thing to do): since my OS is in English even though I'm in France, I get to have the English wikipedia with !w and use !wfr for local stuff. In fact I did not know about the behaviour you describe because of that.

!mdn for developer.mozilla.org is one of my most used ones :)

Yes, that's a great way to skip all those w3schools results. I should probably bookmark the RegExp entry, because I

    !mdn regexp

almost daily.

> The only thing I can't abandon is !gmaps

I was in the same boat with !gi but DDG image search is increasingly relevant, and is not infested by Pinterest results. DDG increasingly feels like home the way Google Search did years ago.

I use !npm which is pretty great if you'd doing node dev