Arfcom engineers will appreciate this story.
I think the course was Dynamics or Thermo - so the math wasn't overly complicated, but understanding the concept and which way forces applied could be difficult.
Well, this professor graded on a set scale, with each question being assigned X points, and percentages of each question's points assigned to various parts of the process, correct work, correct units, etc....
The rest of the numbers don't matter, but I CLEARLY remember that only 10% of the points for each question (and thus the total exam) were assigned to correct assumptions. If your assumptions were wrong, but the work stemming from your assumptions was correct from that point forward, you only lost the 10% of the points for wrong assumptions. The TA would actually trace your (wrong) work and assign partial credit for working the question correctly from the wrong starting point.
Being the loophole-seeking lazy asshole that I am (though I prefer the term "efficient") - I would list out assumptions that were clearly wrong, but that simplified the question to the point of being high-school physics simple. Things like "assume a frictionless environment" or "assume fluid is incompressible" or "assume temperature is absolute zero" or "assume 100% efficiency".
Basically, I use wrong assumptions on purpose to make the questions as simple as possible, and would get 90% on the tests