There are certainly plenty who try using real math, but they run into the limitations that always appear with real math and data: A) Historical data assumes no structural shift (legislation, etc.) outside of the historic range. B) Most tradable things are highly correlated, which has to be accounted for: two similar stock charts are not two independent datasets. Once the correlation is accounted for, even using all global markets gives a fairly small start for large macro forecasts. And people usually prefer an exciting, bold, precise prediction with no grounds at all over a careful mathematical prediction with a full explanation of expedited ranges, significance, and assumptions.I would have thought predictions of market fluctuations without the use of differential equations or other concrete mathematical definitions would be laughed at, in 2016.
All the best equations involve exactly zero people. Well, my favorites, at least.