Market formula. - page 8

New comment
 

It's just the way I did it.

And the 20 bar value, is one of the indicators for the tester to adjust.

Reshetes - 3 indicators, 7 variables for fitting - for example.

The effect is not great if you aim to minimise the stop.

 
alsu:

Mathematically, "there is a memory" corresponds to the statement "there is a difference equation (not necessarily linear) describing the dependence of price expectation at a given moment on previous ones". And theoretically, this equation can be guessed "calculated" based on general considerations about the nature of market behaviour at such moments (i.e. by building a mathematical model of the system).
That's the point, if such an equation exists, it will not necessarily be linear and stationary. And it is possible that one equation is replaced over time by another level and so on to infinity. So "calculating" such an equation will be no better than guessing the next flip of a coin, but will require orders of magnitude more intellectual resources and time.
 

20 bar(in future), RSI, and EMA.

And fitting.

 
And the stop loss is minimum or slightly more ). 10 pips for example.
 
C-4:
That's just it, if such an equation exists, it is not necessarily linear and stationary. And it is possible that one equation is replaced over time by another level and so on to infinity. So "calculating" such an equation will be no better than guessing the next flip of a coin, but will require orders of magnitude more intellectual resources and time.



It's not like anyone suggested calculating it))) The topic starter asked what if there is one. Constant, not changing over time. Based on that, the answers are
 

But it's probably prettier to take a value of 100 bars.

And as a test it should make a profit everywhere, regardless of

the depth of insight. Adjusted.

And then tested from 1 to 100 bars into the forecast.

 
C-4:
That's just it, if such an equation exists, it is not necessarily linear and stationary. And it is possible that one equation is replaced over time by another level and so on to infinity. So "calculating" such an equation will be no better than guessing the next flip of a coin, but will require orders of magnitude more intellectual resources and time.

That is why theoretical work is needed, first of all, to select from many different classes of models those which describe the system (including its non-stationarity) adequately. I agree that this work is, by and large, a scientific poke. I agree that one has to be very lucky (or a bit of a genius) to guess the right model. But I don't agree that it is impossible.
 

in any case, any TS is a probabilistic prediction of future price increments. It is a local market model. The MO of the TS is the MO of the actual price increments over the forecast period. And the variance accordingly. That's why everyone who builds systems deals with such local market formulas.

P.S. the signal to enter is the moment when the MO/dispersion of the forecast is big enough, and the exit (including stop) is when this value is either decreased below a certain value (not necessarily 0), or the further forecast is already unknown

 
Avals:


Many forecasts (price trajectories) are reduced to a probabilistic forecast. You can draw the distribution of the price after a certain period of time, which means you can get a positive mathematical expectation. But this is only if the pricing process is non-Markovian - i.e. has a memory.

A Markovian process is a random process whose evolution after any given value of the time parameter t does not depend on the evolution preceding t, provided that the value of the process at that moment is fixed ("future" of the process does not depend on "past" when "present" is known; another interpretation (Wentzel): "future" of the process depends on "past" only through "present").

If the process is Markovian, then your distribution will always be with mo=0 and change at each step along with the price. I.e. the best prediction will be the current price for any number of steps ahead. That is, you will receive martingale on which you cannot make profit.

The process of price change is non-marting, so your formula is a grail)) Although, it doesn't mean that all trades must be in the plus, because there is variance of future price distribution. The variance is what will mean the error in the prediction. But knowing Mo and variance at any given time you can trade when Mo is large enough and variance is small. I.e. maximize mo/dispersion functional.

The important thing in this interpretation is also the memory depth. And how fast the forecast changes. It depends on it, for how long in the future to predict (how long to keep a position).

The trading system would be simple - enter when mo/dyspersion>X and hold a position until mo/dyspersion>0. I.e. if the forecast has changed and mo is less than zero, then exit :)


There is a clear methodological gap in your description. You should start with a meaningful, verbal description, not a mathematical one. In this case the picture is as follows.

We have a quotient, any, a handicap, an exchange, it does not matter.

In the quotient has a deterministic component, which reflects the inertia of the production processes and services.

Around this deterministic component there is a random, non-Gaussian process, which reflects different cases in production.

Some quotients have a deterministic cyclical component, reflecting the fact that "potatoes grow in summer and do not grow in winter".

Around the deterministic cyclical component there is a non-Gaussian process reflecting climatic randomness

There is a random Gaussian component in the quotient, reflecting the opinion of the crowd in the demand-supply process of pricing.

This all can be modelled but it is not a complete picture, as there are Bernanke, tsunami ... that produce shocks and at these points our deterministic components that have analytical form become inadequate for forecasting. The reason is as follows: after a shock, there are two possible development paths: returning to the initial trajectory and setting a new trajectory and we are unable to predict this choice. We have to wait for a new sample to recalculate the model.

But even this is not the complete picture. Inside the economy there are accumulations invisible in our deterministic components and there are breakpoints, whose occurrence can be known only by having some history. In hindsight we can see that all of our beautiful analytical curves and models of stochastic processes are flying into tatters and new parameters of this described verbal model must be calculated.

 
faa1947:

There is a clear methodological gap in your description. You should start with a meaningful, verbal description, not a mathematical one. In this case the picture is as follows.

We have a quote, any, a handicap, an exchange, it does not matter.

The quotient has a deterministic component, which reflects the inertia of production processes and service provision.

............

I did not pretend to describe pricing completely. I did not describe it at all.) Just what the topic starter was asking about. He did not ask about the pricing model, but how to use it if it already exists.
123456789
New comment