Machine learning in trading: theory, models, practice and algo-trading - page 3206

 
Aleksandr Slavskii #:

1) Tell me what language this is written in and what it sounds like if in human language)


2) Isn't there any way to reduce the amount of data?

Instead of a bunch of candlesticks, you can feed zigzag values. Imho it will describe your pattern on the level no worse.


3) I realise that the whole pattern is much wider than this piece of chart on the screen.

(1) I invented the symantics myself, it's not in any language, just a string that carries information.

This is a mega simple example, more like an illustration of where to dig and where no one digs except me unfortunately....


1 = price above the level

-1 = below the level

0 = price is equal to the level

| = next event


(2) information about candlestick prices is lost, everything is simplified.

(3) everything can be, I showed a simple example, you have already shown a more complex one, one rule alone does not solve anything.

 
Aleksey Nikolayev #:

I also prefer zigzag to candlesticks, but there is a problem of ambiguity with it, because there is a dependence of the resulting pattern on the zigzag parameter. For example, in the first figure, a smaller zigzag may or may not draw two tops instead of one #2, because there are two candles and anything can be inside them.

Of course, you can clearly see by eye what step of the zigzag should be taken in each specific case of pattern implementation, but it is almost impossible to formalise this vision in the form of an algorithm.

Well, yes, there is such a problem with ZZ.

When I was trading, I used nothing but the volume indicator, and now I have to adapt indicators, because I can't write in the code like "well, see this!!!".

I think that ZZ on true fractals (if I remember the name correctly) would be the most convenient, but what is in the kodobase works with errors.

I tried to fix it myself, it almost worked, but still on super large bars, when the fractal is both above and below, it turns out to be some kind of bullshit.

 
mytarmailS #:

(1) I came up with the symantics myself, it is not in any language, just a string that carries information.

(2) the information about candle prices is lost, everything is simplified.

1. Thank you.

2. So if everything is simplified, that's a good thing. Or am I misunderstanding something?

Why do we need information about candlestick prices if

The main task is to determine the direction of price movement online. The simpler it is, the better.

 
Aleksandr Slavskii #:

1. Thank you.

2. So if it's simplified, that's a good thing. Or am I misunderstanding something?

Why do we need information on candle prices

The main task is to determine the direction of price movement online. The easier it is to do this, the better.

2. If you simplify to remove unnecessary information, it is very good, if you simplify for the sake of simplification and lose important information, it is bad.
 
mytarmailS #:

A level is a price around and within which important events occur (breakout, breakdown, etc.)

It is important not only that important events occur around the level, but also the patterns of events themselves (i.e. it is important not only that the level was broken, but also how exactly it was broken... or rebounded).


An example of description of one of many levels.

The blue circle indicates the place where the rule was triggered, after the circle the data is not available for the algorithm.

This is all one and the same level, which is described by this rule


Is there any chance to describe this non-linear event by inputting the last 5 candles????

Is there any chance to find such levels by comparing them by some metrics, let's say through Euclidean distance????


Of course not.

That's where the level is most likely to be.

where your circle is, it is most likely a test, i.e. a check by the quoter for worse behaviour of some trader.


 
Aleksey Nikolayev #:

Of course, you can clearly see by eye which zigzag step should be taken in each particular case of pattern implementation, but it is almost impossible to formalise this vision in the form of an algorithm.

Probably, there are auto-tuners of the ZZ.

 

If anyone wants to test this in a tester

Initial data:

10-period MA at the last 55 closing prices minus these closing prices. On M15 TF.

Reference pattern:

[-0.000538, -0.000137, -0.000404, -0.000396, -0.000544,  0.000102,  0.000556, 
  0.000215,  0.000247,  0.000163,  0.00025,  -0.000006, -0.000045,  0.000322, 
  0.000375,  0.00008,  -0.000269, -0.000095, -0.000509, -0.000458, -0.000412, 
 -0.000472, -0.000532, -0.000608, -0.000323, -0.000575, -0.000459,  0.000302, 
  0.00008,  -0.000118, -0.000195,  0.,       -0.000833, -0.000845, -0.000353, 
 -0.000455, -0.000855, -0.000683, -0.000607, -0.001006, -0.000945,  0.000019, 
 -0.000518, -0.000373, -0.0004,   -0.000336, -0.000442,  0.000183, -0.000087, 
 -0.000167,  0.000019,  0.000727,  0.007636,  0.007539,  0.006659]
 
Aleksey Vyazmikin #:

So misunderstood past posts.

So, if you approach the problem not head-on, but more creatively.....

You practice averaging, imagine that the trend is a monster, and opening a position is a shot at it, for neutralising (closing in the plus) the monster with a smaller number of shots to give the greatest reward. The volume of a position can be thought of as an energy charge that is decreasing, and the more of it retained the better. Trends are independent monsters for training.

Wouldn't you rather write a new chapter of Harry Potter?

 
Maxim Dmitrievsky #:

Why don't you write a new chapter of Harry Potter?

And you're a fan if you have to?

 
Maxim Dmitrievsky #:

If anyone wants to test this in a tester

Initial data:

10-period MA at the last 55 closing prices minus these closing prices. On M15 TF.

Reference pattern:

I, I, I want it. I just don't know what it is and how to check it.

I tried to read this thread first, so as not to ask unnecessary questions, but it turned out to be problematic because of the water, which in this topic is much more than any little useful information.

1. What is this "pattern explorer"?

2. How to check?

3. what is a " Reference pattern"? Where does it come from?

4. What do the lines shifted upwards at x=57 (black vertical line) tell us ?

5. The 10-period MA is one, where did the 21 pieces of these mashas come from ? or not mashas, what is drawn on your chart ?

I'm trying to get to grips with the MO, but so far I haven't made any progress. There are a lot of articles written, but mostly for those who already know the basics, but there are few of them (I haven't found them yet).

I tried to read your article "Fuzzy logic in trading strategies", but after reading this:

"Fuzzification of input variables. Establishing a correspondence between the numerical value of the input variable of a fuzzy inference system and the value of the membership function of the corresponding term of the linguistic variable."

My brain snapped.

In my youth I used about the same words when I was picking up girls )

I understand that it is not easy to write an article, and it is almost impossible to write in such a way that you are understood, but still we should not forget that an article is not only an expression of your thought/idea, but first of all it is a teaching material, so we should write as simple as possible.

This is not even a criticism, it is a request, if you do decide to answer my questions, I would like to see answers either without specific terms or with an explanation of them.

Reason: