Some doubts regarding jaro_winkler_similarity and jaro_similarity results

[passcombo]

V 11.0 Tested for simple a/ab combinations, does not sem intuitive are result jaro_winkler_similarity ('a','ab') = 0.85 jaro_similarity ('a','ab') = 0.83

according to wiki 1/3 of sum should be about 0.5 ?

https://en.wikipedia.org/wiki/Jaro%E2%80%93Winkler_distance

jaro_winkler_similarity ('a','ba') = 0.0 jaro_similarity ('a','ba') = 0.0

same here should be about 0.5 ?

Or maybe I confused something?

[jamesturk]

Hi,

I'm not sure I understand your question or what you're doing to get your results where the similarity equals zero. Or why it would equal 0.5.

As for the first part, the equation on Wikipedia shows that

This means 1/3*(1/2+1+1) which is 0.83333 -- the winkler modification accounts for the boost to 0.85.

[passcombo]

Thanks, I counted the transposition wrongly. But in any case the other example with reversed order "a" vs "ba" gives zero, which indicates the method jaro_winkler_similarity has some limitations and the other method - damerau_levenshtein_distance - seems to work better for me.