If I just suggest a reference without a suggested change in "point of view" you're likely to get very confused.
May I recommend as a starting point:
Be skeptical about the idea of a single number characterizing intelligence. We don't even do that for the weather.
Next, learn about the genesis of the entire IQ concept, and later efforts to academically justify past ideas about intelligence. The first really popular writings were by Galton
http://galton.org/books/hereditary-genius/text/pdf/galton-18...
Afterwards, Binet and Simon, and Terman (at Stanford, the Stanford in Stanford Binet) get at it.
Back then, IQ was really a "quotient," "mental age" divided by actual age. The concept only really makes sense in a developmental context -- after all, how are we supposed to believe that a "developmentally baseline" 23 year old is actually the pinnacle of intelligence?
Only after this did the Weschler Adult Intelligence scale come in. Read (some) of the criticisms Weschler had of the Stanford Binet here.
https://en.wikipedia.org/wiki/Wechsler_Adult_Intelligence_Sc...
This was devised as a clinical scale. This is still one of the only widely administered and normed tests. The norming takes place only over a few thousand individuals. The types of abilities typical in an international IMO context are simply unlikely to show up in these cases, much less register statistically in the test!
It's also important to learn about the hypothesis behind "Spearman's g", https://en.wikipedia.org/wiki/G_factor_(psychometrics)
and the mathematics behind what suggested it, "Factor Analysis," which has many capabilities but is worthy of many criticisms, including particularly that there is no obvious weigh to norm or weight any of the tests, so ultimately you end up with a judgement call for which tests to include, and how much to gather from their results.
A far more enlightening study of genius, I have found in the book "Human Accomplishment", by Charles Murray. When you're sample is all of those humans who have left a mark on history, and your measure is impact, and our remembrance of it, you learn far more about the heights of human capability, I believe, than by scoring how well people match patterns to a preconceived idea of the "right answer" on a test.