>>11337482Well, the thing is that professors know homework in lower level math classes is generally not indicative of skill or mastery due to all the answers being standard, etc etc.. So the exams generally do the heavy lifting to separate people who understand from people who don’t. At higher levels, it becomes more and more common for professors to either write their own problems or to draw them from a myriad of different texts so honing down locations for answers end up funneling you through resources that actually just help you learn the material.
For example, for analysis, we would have like 5 problems with 2-3 parts each. The first two problems would be easy original problem / an easier Rudin problem. The next two problems would be more complex and moderately harder in principle to solve, and they were usually substantially modified versions of problems from other texts. The last problem, which was always the hardest and would take ~5-6 hours over a few days, would generally be either completely original or a heavily modified version of an introductory grad problem, so any search for answers would only leave you with ribbons of intuition from math exchange and not much else.
It was much more efficient to solve out the problems in a large study group, and that was the professor’s intention.