People seeing this problem are correct… TMDB seems to be messing around with their already kind of hoky search algorithm. I too lost the ability to search correctly for quite a while, and then suddenly it came back. (While the whole time the movies were fine when searching directly on the web-site.)
The algoritm they use seems a little strange to me. It at times seem to have a hard time finding content without modification. For example, for most movies names that start with a single "A " (as in “A dude called Riddle”) even in the best of workings I usually have to delete the "A ", and then the movie is found. As I sync with a Linux Samba server, I use “_” (underscores) in my folder names. Most of the time, TMDB correctly parses these out, but sometimes I have to manually remove one or more of them using the keyboard function for a search to complete.
Except in the case mentioned by the other writers, where everything suddenly stopped working (only from the Hub, not from the web-site) and almost nothing was found. During that time, I at times had success with removing the whole movie name except for the first word or two, searching on just that prefix and then scrolling through the much longer selection TMDB then might find. But it was surely a pain until it reverted itself.
BTW. I share this exact same directory and file structure with MediaCenter with an add-on from My Movies, which uses its own movie database for lookup. My Movies auto-searches for movie information when it detects a new directoty, and finds the correct movie probably 99.9% of the time, based on the exact same folder names that TMDB have regular trouble with, so I have to manually edit searches in place.) I then just have to click OK to accept what it found, all by its lonesome.
“My Movies” is much more complete than TMDB, and actually is more than a movie database, as it contains all DVD and release versions separately, similar to IMDB. This I do not care too much about. The main thing for TMDB to learn from is the pattern matching algorithm used. It seems much, much more accurate, and separates words both by under-score, space, Uppercase letters, and other.
Plus it can handle when a movie name starts with "A " (An A with a space after it) 