Why does a wildcard search bring back fewer results than a non-wildcard search?

[mlissner]

My search for euphemism ...

https://www.courtlistener.com/?q=euphemism&type=r&order_by=score%20desc

...got 1227 cases, while one for euphemis* ...

https://www.courtlistener.com/?q=euphemis*&type=r&order_by=score%20desc

...got only 895.

I would think the latter should yield a superset of the former, no? What am I missing?

[mlissner]

Tricky one, but this has to do with stemming, which is the way search engines take off the ends of words to make them into base words. This way, for example, if you query for run, you'll get results for running, runner, etc, but probably not ran, because stemmers don't understand grammar, and only cut off word endings. If you want the details, or to try playing with some words, we use the English language snowball stemmer, but there are different stemmers that have been tried over the years and lots of them for different languages too.

To get back at your question, the answer is that when you use a wildcard, it disables the stemmer. So the first query, euphemism will actually do more aggressive matching, returning words like, eupheme, while the wildcard search will not, since it's only looking for words that start with euphemis.

Stemmers are sometimes controversial because of this kind of thing, but usually it's a better balance to return more results than fewer and to not make users do queries like runner OR running OR ....