As an undergraduate (I don't study maths primarily, but employ these in the study of econometrics/economics), I've come to realise that solving difficult problems in fact hinge on one's ability to recognise relationships between two things, and then (though not always) expressing this relationship in some equation which moves one towards a solution. This relationship to be recognised will be one which was previously not recognised, or at least not explicitly covered in class.
My question therefore, is how can I get better at this skill at recognising relationships? I'm glad that I've pinpointed the skill on which solving difficult problems hinge (or at least have a working theory), but I still fall short when actually solving difficult problems in problem sets and exams.
For example, suppose in an exam I am asked to show that the expectation of a non-negative random variable is equivalent to the integral of its complementary CDF from $0$ to $\infty$, that is, $E[X]$ = $\int_0^\infty P(X >x)dx$, where $X$ is a non-negative random variable. I had previously only learnt the basics of probability distribution function, cumulative distribution function, expectation and variance etc., but had never had to connect the concepts of CDF and expectation. Mike Spivey, in the accepted/best answer to this post, illustrated why this proof holds true, and, to use the language I employed previously, I note that his explanation hinged on recognising the relationship between column addition and row addition. This is seen from his sentence, "we could also add up these numbers row-by-row, though, and get the same result."
This is of course but one example - my question is how do I train myself to be able to recognise such relationships, especially under the time constraint of examinations, so that I may solve difficult questions? Is there a systematic way to go about doing this?