I don't see that there's a better way to divine the truth in historical, economic, and political matters.

I didn't ask for a full statistical analysis. Although it would certainly be nice, it seems like a lot of work - deductive reasoning is but a tiny subset of probabilistic reasoning. However, I suppose one could use random sampling methods to make sure that you didn't go far wrong. New field name: asymptotically correct history.

Anyhow, what I said was lacking was evidence in general - arguments that are harder to make when false than when true. This sort of thing often does mean statistics, but not because statistics are magical - because when you use statistics it's easy to see when you leave stuff out.

I find that a sharp mind - Moldbug's is extremely sharp indeed - is capable of achieving a high signal-to-noise ratio using this style of reasoning where others might not.

Correct reasoning is not a magical power, it's a process that anyone can follow. A->B to ~B->~A, non-replicable reasoning is highly suspicious.

After all, a direct confrontation of the reasoning style you disparage and an approach I expect you might consider more "evidence-based" is found in macroeconomics

If we drag economics into this we'll be here all day :P Anyhow, I'd rather pick an example with representation from the "literary" side, and maybe a control group. Unfortunately, no clean examples come to mind.

Arguing from principles and representative quotations is a good way to give evidence for claims about the way people say things

I can only interpret this as the idea that all speech and writing provides evidence only about the speaker or author himself. I disagree in the strongest terms!

Please interpret it literally.