An area of maths your touching on is called, curve matching, pattern matching, over optimisation, data fitting, it even touches on statistical significance.
Basically you can take any complex set of data and find patterns in it, but 99.999999% of the time they have absolutely no relevance. This is a common problem with newbie’s to financial markets; they will take 20 years of data, and mine it for patterns. IE a newcomer may ask, what is the best day of the week to go long (as a silly example)? And so they crunch their 20 years of data and work out Monday is the best day to go long. However, when they start trading this strategy they can’t work out why they are always loosing.
The reason lies in the fact, that all they have done is over optimised a set of random data, sure in 20 years Monday was the best day to go long, but if you looked at 40 years Tuesday was the best, and over a 100 years Wednesday was the best.
There are many examples of how this can work with or without financial data. Another “loose” example would be someone who looks for apparitions. The image being complex data, the face in the image being over optimisation. For if the pictures was taken at a different time of day, different angle, or even with a different camera, there would not be a face, or another (different face) may appear. It’s also easy to see there are sociological effects going on in my last example, but the theory is still the same.