From theory to practice - page 533

 
A simple mashka and a curvilinear line built by the midpoints of the linear MNK segments are one and the same. Well, that is, we take for example a linear MNK with 11 points, build it, memorize the 6th point, move it forward by one point, build the MNK with 11 points again, memorize the 6th point and so on through the whole row. The result is exactly the machka. Curvulina is a linear variant of the Savitsky-Haley filter for 11 points. In spectrography for approximation sometimes non-linear variants (cubic or even higher degrees) of this filter are used, for data preprocessing.
 
sibirqk:
A simple mashup and a curvilinear segment based on the midpoints of a linear MNC are the same thing.

Who uses MNCs there?

An ISC is, after all, taken over several points, not every two points a different segment... Not to mention that MNCs can be taken to a higher degree - second or third.

 
Yuriy Asaulenko:

Well, that's what modelling is for, not to write a PBX straight away.

I modelled my previous PBX for about six months. It ended up being a very simple and beautiful system. But in the process it was more than enough complications).

Here is almost no visualization. All logs are written in an Access database.

I would like to agree, but ... But I want to argue ... ))

you need to formalize the task, on elementary (terribly simple!) tasks:

- the result of writing an ATC is what - to get a balance chart!

- ideally, we want a balance line at 45 degrees upwards, to aim for such a line, we need to know

a) the forecast of the movement

b) the reliability of the forecast

i.e. by deep reasoning, we arrive at the video:https://www.mql5.com/ru/forum/221552/page525#comment_8564120

task a) : what do we need to predict? we need a price movement pattern:

- trends = line

- channels = 2 lines

- levels = ??? but I want to predict this, I suspect that it is a few horizontal lines which will change their values depending on the time of holding the position in the market

task b) is solved by the strategy tester and statistical observations

that's about it, the minimum left to do is to deal with task a)

От теории к практике
От теории к практике
  • 2018.09.03
  • www.mql5.com
Добрый вечер, уважаемые трейдеры! Решил было на какое-то время покинуть форум, и сразу как-то скучно стало:)))) А просто читать, увы - неинтересно...
 
Igor Makanu:

I wish I could agree, but... but I want to argue... ))

you need to formalise the task, into elementary (terribly simple!) tasks:

- the result of writing an ATS is what - to get a balance chart!

- ideally, we want a balance line at 45 degrees upwards, in order to aim for such a line, we need to know

a) the forecast of the movement

b) the reliability of the forecast

i.e. by deep reasoning, we arrive at the video:https://www.mql5.com/ru/forum/221552/page525#comment_8564120

task a) : what do we need to predict? we need a price movement pattern:

- trends = line

- channels = 2 lines

- levels = ??? but I want to predict this, I suspect that it is a few horizontal lines which will change their values depending on the time of holding the position in the market

b) is solved by the strategy tester and statistical observations

like this, the minimum is left to solve the task a)

Rejected:

- the model of price movement (better to say, almost rejected),

- I gave up forecasting (almost too). Everything comes from statistics and probabilities. I.e. losing trades are already an initial property of the system.

- I have given up optimizing for maximum profit. As a result we get a lot of different criteria, but all subsystems are set up mainly separately (independently).

 
RRR5:

well I'm using polynomial, so it's "different parts of a parabola".

You could also try using an arc, but I don't know how to do a regression on it.



Can you derive the arc equation from here to score it in excel?

http://sci.sernam.ru/book_geom.php?id=39

Thanks for the link, got carried away looking for related ideas. I don't think you need to approach the question as there, from the perspective of R-functions. Approach with conventional means:

Directly from the equation of a circle of radius R with centre coordinates Xc, Yc, ignoring the loss of meaning in units (number, course, distance)

R^2 = (X-Xc)^2 + (Y-Yc)^2

we make a function of the inconsistency at the point with the number i (Xi = i). This is the difference between the distance from point (Xi, Yi) to centre Xc, Yc and radius R:

Di = ((Xi-Xc)^2 + (Yi-Yc)^2)^0.5 - R.

The squares of Di and add up to obtain the target function to be minimized. There are three parameters to be varied: R, Xc, Yc. At extreme (first and last) points the influence is less than at intermediate points (no adjacent), I think it is better to multiply corresponding Di^2 by two. And for your fitting purposes with emphasis on the last points, you can also play with weights of several last points by setting up one more column for weights.

If it is not straight, the units of measurement will have to be taken into account. In order to make X and Y influence almost equal in distance calculation (and R is affected by both), it's necessary to take as X not i number, but the same number multiplied by alignment scale, so that X and Y ranges are close in size.

P.S. It turns out that sernam.ru has very cleverly got rid of the accusations of copyright infringement by publishing the texts of books only in parts and without specifying the output data, in particular the titles of the books. It is possible to find texts on sernam.ru which can be found nowhere else on the internet.

 

I like to work on small TFs, but I don't like these snafus.

How do you predict them?

GBPJPYM1

 
Uladzimir Izerski:

I like to work on small TFs, but I don't like these snafus.

How do you predict them?


Why "predict" them - they need to be "handled" correctly! There is nothing special about them, they need to be "detected" in time and properly "handled".

 
Uladzimir Izerski:

I like to work on small TFs, but I don't like these snafus.

How do you predict them?


There's no such thing.

because absolutely

all mothermothermatter averages

 
aleger:

Why "predict" them - they need to be properly "managed"! There is nothing special about them, they have to be "identified" in time and properly "handled".

Detecting is always too late. But predicting is just that.

 
Uladzimir Izerski:

Determining is always too late. And predicting is just that.

If predicted verbatim, then yes, no problem, anything. But justifiably, that's unlikely. "Determine" is not always to be late, on the contrary, it is more to be "ahead"!