if you treat duplicates as a set you can remove the second condition because if you see a repeated value, the size of the set remains the same instead of skipping over it (your solution).

the .trim() method was clever. I just did the last line manually.

This seems inefficient imo, because you everytime you make the recursive call you are starting over at the first pair. However, it doesn't make calls on the incorrect buckets.

I don't see where it creates a new string. Can you elaborate?

Can someone explain why he processes the roman numeral backwards?

Why is this so overly complicated?

Is it reasonable to assume the possibe inputs by taking a look at the sample teset cases? Thank you!