An interesting concept.
The fact that punctuation is "typed" automatically is a little jarring, especially since the spaces before or after any punctuation is NOT automatically counted. I imagine if a person played it long enough, this might not be an issue.
But if the idea behind the whole thing is to type what you see accurately as quickly as possible (and it is), then automatically counting punctuation but omitting spaces seems just plain inconvenient.
Does that make any sense? I'd love to see another version of this where everything you'd have to type in a real essay acted as a barrier to your progress.