I was wondering this as I read it as well. Yes, we know it would be fairly easy to compromise an individual account by guessing common passwords. But that assumes you have one particular account in mind. If you're just looking for bank accounts to access, you first need a list of
usernames to try (ideally usernames of non-tech-savvy people) and then you need to run your entire password guessing routine for each one.
I'm no security expert, but I start to wonder if the "practical" entropy of a given password is actually much higher than the theoretical entropy numbers people derive.
Serious question: when is the last time anyone has had an account compromised because of a weak password?