Statistical arbitration - page 5

[Alexey]

hrenfx:
You should not watch the video, but move the construction interval yourself with the mouse. Fortunately, the toolkit immediately (without lags) rebuilds the synthetic, showing it beyond the construction interval (red triangles - first zero crossing by the synthetic outside the construction interval).

Well, the only thing left to do is to find out how to attach the indicator from MT4 to the Expert Advisor on MT5. I personally find it difficult to understand the logic of the program without comments. But it is nicely done, no one argues.

[Heroix]

ivandurak:
The only thing left to do is to find out how to attach the indicator from MT4 to MT5 EA. I personally find it difficult to understand the program logic without comments. But it is well done, no one argues.

I've got a crude version of the Recycle converted to MT5 somewhere... I will have a look at it tonight.

[Alexey]

Heroix:

I've got a crude version of the Recycle converted to MT5 somewhere... I'll have a look in the evening.

You don't need to, the supposed variant should be able to match currency participation coefficients to any requity set by the user. If you want to help, come up with a formula to characterise requity with one number. For example by maximum balance, minimum drawdown. As long as the degree of similarity to a straight line is assumed. The ratio of the slope of the linear regression of the aggregate equities to the variance relative to that regression, the larger the value, the better. This is needed for the genetic algorithm for fitting the coefficients.

[hrenfx]

ivandurak:
 в предполагаемом варианте должны подбираться коэффициенты участия валют под любую иквити задаваемую пользователем.

Hence:

One can fit coefficients on any window to any fictitious mathematical condition - a kind of "pulling the market on formulas". Therefore, one should not approach this process from what kind of synthetic graph one would like to get, but from the basics - the search for market correlations.

ivandurak:
The ratio of the slope of the slope of the linear regression of the aggregate equi...

It is incorrect to immediately - (abcissa - time, ordinate - money) .

ivandurak:

This is needed for the genetic algorithm for fitting the coefficients.

Without an analytical solution, it will be extremely difficult (long or expensive) to recalculate the floating window (movement of the plotting interval) to test the robustness of the approach.

[Alexey]

hrenfx:

Hence:

Incorrect immediately - (abcissa is time, ordinate is money) .

Without an analytical solution it will be extremely difficult (long or expensive) to recalculate the floating window (movement of the build interval) to test the robustness of the approach.

Personally, I have slightly different ideas for portfolio trading. I am not interested in the stationarity of total equity, I am interested in another thing - whether the weights remain relevant until the signal for closing a position is received, and besides the value of each instrument decreases when the number of instruments increases, which must decrease the role of stationarity for each individual instrument.

As for incorrectness, this approach of Equity Analysis allowed me to build an EA on two wagons which forward testing exceeded the optimization period 10 times without losing money, I'd better not say anything about drawdowns.

As for formulas and pulling on them, and if by some universal law a sinusoid will be better for earnings. Conversion to a straight sine the hypotenuse by idea should be the matter of five minutes. Although of course I understand that the direction is most likely a dead end.

As for the analytical solution. Do discount my nickname, I took it for a reason.

[hrenfx]

ivandurak:

Personally, I have slightly different ideas about portfolio trading. I am not interested in the stationarity of total equity, I am interested in another thing - whether the weights remain relevant until the signal to close the position is received, in addition, when the number of instruments increases, the value of each individual instrument decreases, which should decrease the role of stationarity for each individual instrument.

As for stationarity I stated on the same link above.

ivandurak:
The ratio of the slope angle of the linear regression of the total equities to the dispersion relative to this regression, the larger the value, the better.

The idea is clear.

Here we get a serious dependence of the result on the interval of the linear regression and the way of quantization of the qVR. I.e. everything is rather ambiguous.

Perhaps the following criterion will come in handy:

1. Construct a triangular matrix A[i][j] - number of knees >= i pips inside the outermost knee >= j pips.
2. The sum of matrix elements is the optimization criterion. The smaller it is, the better.

[Alexey]

hrenfx:

About the stationarity, the same link I gave above is all there is to it.

The idea is clear.

There is a serious dependence of the result on the linear regression interval and the way of quantisation of the qVR. I.e. everything is rather ambiguous.

Perhaps the following criterion will come in handy:

1. Construct a triangular matrix A[i][j] - number of knees >= i pips inside the outermost knee >= j pips.
2. The sum of matrix elements is the optimization criterion. The smaller it is, the better.

The interval, instruments on which the ideal equities are built, are selected by the user. More details on the optimal matrix, better with pictures. A preliminary outline of the indicator is in the trailer. The wishes are accepted for now.

[hrenfx]

1. To make the variance of the linear regression depend only on the way the qVR is quantized, put a constraint on the changes of the weighting coefficients. For example, Sum(Abs(K[i])) = 1.
2. You can completely forget about TickValue if you work with logarithms of prices.
3. About the matrix by example. See the outermost knee of the ZigZag whose min knee condition is 100 pips. Calculate inside it the number of ZigZag knees with the condition of a min knee of 10 pips. Write the obtained number in A[100][10].

[Alexey]

This is the interesting thing about it. We calculate the optimal suitcase for the period between the green and red lines. Using the participation coefficients of currencies in the suitcase their aggregate Equity is calculated. I cannot use the indicator properly. And one more question, who knows how to delete graphical objects when deleting an indicator.

[Anatoli Kazharski]

ivandurak:

...

And another question, who knows how to delete graphical objects when deleting an indicator.

In OnDeinit() you simply delete the required objects.