@freemo yeah I'm not a huge fan of the error-catching process either. But it's at least measuring the same way across the board, whether that's by tokenizing words or a quarter of the letter rate or whatever. It's easier to interpret a number where I have that context than one like your keyboard's self-measurement where I don't.
I just did five and got 88, 76, 88, 80, 93. Your 127 would have been first in any of those races.