85% is surprisingly high for fingerprinting, hence self-deprecation over insulting the author by poking at efficacy. I wouldn't have expected my Australian spelling, Oxford comma, or cadence to be anything close to the Californian Rust enthusiasts I apparently match against. Especially as there's no normalization happening - so even the Burrows-Delta method shouldn't match my use of "gaol" or "humour" that often.

But, limiting to the top couple hundred words, probably does limit me to sounding like a pretentious dickhole, as I often use "however", "but", and "isn't". Corrections are a little too frequent in my post history.

I'd expect things might be a tiny bit looser with precisions if something small like stop words were removed. Though, it'd be interesting to do the opposite. If you were only measuring stopwords, would that show a unique cadence?