Does this "reduce draws" purely because it randomly gives one side a small (or big) advantage? If so, that doesn't seem to be an improvement at all.

It would be easy enough to test if one side having an advantage is common: have a chess engine play against itself many times using the same setup. The try a new setup. Find out what percentage of the time one side seems to have been given an advantage. (i.e. an advantage that's greater than the built-in one that white has.)

If there's a different reason this reduces draws, what is that reason? If it's simply reducing the reliance on book openings, that's fine, although there are other solutions for that. If there's something in a particular piece structure that reduces draws, what is it?