I'm interested in your initial idea, but there has also been lots of thought provoking commentary. Perhaps this is a naive question, but why has no one made a strategy to test the initial theory? It seems like it could settle some of the disagreements if you run it on multiple instruments/years. I tried to make one myself, but got stumped on a few aspects (beginner to NT7).
In my experience, advanced mathematics does help in trading if "advanced" means robust and careful, not super-complex. Some of the relatively recent advances in data mining deal with various forms of regularization. It allows one to reduce the chance of missing structural change in the market at the expense of a fraction of Sharpe ratio. Examples of useful methods are LAR, lasso, nonparametric association measures, etc. But this is a long journey. To understand the methods on the level of their inner workings, to be able to combine them efficiently, one has to invest much time. Something along the lines of this statistics study plan (page bottom, skip the beginning) or a similar kind. Well, 25% of that should do as a start... Advanced mathematics is not binomial distribution or pure Monte Carlo.
The following user says Thank You to staassis for this post: