There are maybe 10 to 15 ways the normalization doesn't manage to do what it's supposed to. Certain committers happen to use profanity more, for instance. Commit style and frequency. Tendency of programmers of language X to be acculturated somewhere that dampens their natural tendency to curse when frustrated.

For instance, my style of coding means I branch off, commit early and often, with flippant and meaningless messages, and then I squash and rewrite my entire change as a single, formal entity before releasing it for coworker consumption. I curse up a storm in my commit messages, but you won't see any of them unless you look inside my computer before I delete the working branch.

This is fun and funny. Nothing about it even starts approaching validity except for one statistic: "210 out of 929857", and even that gets shot down. The entire thing would have been better if he had included "hate" as a curse word.

I think you've forgotten what was being discussed, because your objection isn't relevant. The claim in dispute is: "he does not normalize the graph relative to the total amount of code in each language".