As seen at AstroNerdBoy, there's another manga-based quazi-textbook. His key question is, "So, will this work? Will people who read this actually learn statistics?" The answer to the complimetary inversion is obvious: those who read in order to learn will not read this. Certainly there must be better books for difficult subjects.

That said, I came across a beautiful example of this approach: a book by some French guy, called "Radio? It's very simple". It was an Abbot and Costello dialogue about the vacuum tube RF technology without the associated formulas. I wish an English translation existed. It certainly was entertaining to read, and it delivered the subject as effectively as Horowitz and Hill. I don't know how amenable statistics is to this kind of take. It sounds like something that cannot be taught without formulas.

UPDATE - Sixten e-mails:

Even if you don't learn statistics from this, it may make you excited enough about it that you would want to learn it yourself with other materials. It happened with me and Hikaru no Go.

It's the standard line of argument for this kind of materials, but I am not sure how pronounced the effect is, statistically (please pardon the pun).

He also links to Mathematical Girls.

STEVEN points to Larry Gonick. I'll check it out later.

BTW, let's not forget such examples as "萌え萌えうにっくす"/"Moe-moe UNIX" (ISBN 978-4839909550) and "ツンデレ★りなっくす"/"Tsundere*Linux" (ISBN 978-4896273021), the latter profiled at Akibablog (NSFW ads).