After further analysing folly code, I am quite confused by where that 30x times speedup claim comes from.
If I can see the code right, the about only situation where any speedup compared to brute force happens is when the last character of string is matched, which is perhaps important for some corner cases, but for general case it is virtually useless...
(In contrast, _proper_ Boyer-Moore skips almost always, but well, that requires prebuilding those skip tables...)