Theorem on the intersection of two MAs - page 2

 
Neutron писал(а) >>

...

Thus, optimization of such TS on historical data by your proposed algorithm will only reveal harmonic with maximal amplitude. And everything would be fine, but for one BUT - the position of such a harmonic is not stationary in principle. Therefore, to build a profitable TS using two muxes crossing is impossible - the optimization parameter is not stationary.

...

Yes, the position of the harmonic is not stationary. But the point is (in p2.) that it is always possible to choose an area where optimization can be done, i.e. the position of the harmonic is (almost) stationary.

Or is it NOT always? That is the question.

 
Itso писал(а) >>
IMHO the problem is that a number of these optimal parameters for each period is random.

You mean the optimum parameters do not lie on a smooth curve, but there are gaps from period to period? Quite possibly. But again the question is what the set of optimal parameters is - if they lie on the continuous curve or on the piecewise-continuous one, or if they are single points.

Maybe they may lie on a smooth curve for another indicator.

 
diakin писал(а) >>

You mean the optimum parameters do not lie on a smooth curve, but there are gaps from period to period? Quite possibly. But again the question is what the set of optimal parameters is - if they lie on the continuous curve or on the piecewise-continuous one, or if they are single points.

Maybe, they may lie on a smooth curve for another indicator.

If there is no spread, you can always do it.

If there is a spread, with a large sample you cannot.

Maybe an example of a simple Expert Advisor on two mouves, and test it on the history?

 
diakin писал(а) >>

Yes, the position of the harmonic is not stationary. But the point is (in p2.) that it is always possible to choose a section where optimization can be done, i.e. the harmonic position is (almost) stationary.

Or is it NOT always? That is the question.

Non-stationarity implies non-repeatability of the result, and in the future as well.

Generally speaking, in this formulation, the tading problem is reduced to a search for a stationary parameter characterizing the initial BP or its derivatives. It is definitely not a harmonic BP spectrum.

 
Let's look at the other side:
Is it possible to come up with a BP in which a system of two MAs would not be profitable at the final spread?
- An example of such a segment is a long gently sloping slope with superimposed double amplitude oscillations up to 3x spread.
no combination of MAs is catching.
 

In general, the meaning (or one of the meanings) of the question was whether the optimal parameters of a mask or other indicator on a given interval can be calculated directly from the price series, without carrying out optimisation.

 
You mean from a theoretical point of view?
 
resuscitating the subject of hairiness
 
What are you talking about?