It would be nice to consider how auto-completion actually biases the real distribution of key-presses, e.g. I wouldn't expect closing parenthesis or brackets to be pressed as often as their opening counterparts.