The only time theoretical modelling happens in most civil engineers careers is at university. Especially so with geotech/soil mechanics - it is the most inexact and variable/unpredictable area especially when water gets involved.

Structural steel would be the most predictable. Concrete and timber are in the middle somewhere.

>As they say, the rules are written in blood

Basically nobody ever died from leaky pipes or substandard weatherproofing. The code is as much about a) homogenizing the industry so big business can statistically reason about it at scale b) turning the subjective into the quantitive so that things can be done, checked, sight off on, etc, etc, without anyone using "judgement" as it is about protecting life and limb. Just about every professional has a laundry list of complaints about their area of code that boil down to it being theoretically useful but at great "not worth it" expense or a similar "not worth it" expense being incurred in lieu of very basic judgement. Arc fault breakers, and engineering requirements for small retaining walls come to mind as oft cited examples. And of course there's the myriad of wrangling that goes on wherein things get looser/stiffer requirements depending on whether their use is deemed worth incentivizing (this stuff usually lives in local addendums to the code).

I'm not saying there isn't value in there, but this habit people have of acting like it's all relevant to safety and screeching about "written in blood" is exactly what creates room for unrelated stuff to exist in the code.

>However, when you start reading, it is clear that much of the work is empirical, heavily localized and based on a great deal on the experience of the builder. I found very little in the way of solid theoretical modeling, but lots of measure, adjust, etc.

Which is a point very much in favor of the amateur.