Machine learning in trading: theory, models, practice and algo-trading - page 1712

 
mytarmailS:

Well that's fine, so I just thought you were uncomfortable with catbust.

Yes, but thanks for the alternatives).
 
elibrarius:
Yes, but for alternative solutions - thank you).

You're welcome))


The book is great, I recommend it... I especially liked the prophet section, I thought it was a rattle, but it turned out to be a serious instrument.

 

When scientists want to understand a complex process, they try to decompose it into simpler components and analyze them; spectral analysis was created for this purpose. Let's try to play scientists) even though they're not very successful. I figured out how to decompose the price into simpler components. My decomposition has no additivity, and that's bad, but it's still interesting to look at the price from a different angle .

so, we need the closing price and volatility (high-low)

Let's turn the price into a conditional binary form - if the price increment is higher than the previous one, then "1" if it is lower, then "-1".

code on R

close.bin <- c(0,diff(close)) 
close.bin[close.bin>=0]  <- 1
close.bin[close.bin<0] <-  -1

we get binary price

you can make it cumulative and compare it with the price.

close.bin.cum <- cumsum(close.bin)

It does not look like much) And now let's add volatility to our series

volatil <- high-low
vol_clos <- close.bin*volatil
vol_clos.cum <- cumsum(vol_clos)

That's better...

Ideas...

IDEA 1

So, it turns out that almost all the "weather" is made by the "intra-schedule" volatility, instead of the "binary" price direction. The point is that volatility has a pronounced seasonality and is relatively easy to forecast, the only thing left is to forecast a binary price, which is simpler in structure than a normal one, and then we simply put all forecasts together and get a full-fledged forecast...


IDEA 2.

All proper MO algorithms learn badly from raw prices, even normalized ones, because they don't have repetitiveness in series, probably just because of volatility, which is constantly different, if we decompose the price into binary and volatility and normalize volatility and add them back, or do not normalize but feed them to MO, in theory we should get better generalization ability because repeatability will increase


IDEA 3.

With decomposition we can smooth out prices without losing latency. You can decompose the price and interpolate (stretch) the volatility and the price separately and then add them back together

close.bin <- c(0,diff(close)) 
close.bin[close.bin>=0]  <- 1
close.bin[close.bin<0] <-  -1

volatil <- high-low

#  растянем цену с 50 точек к 300

cl.big <- approx(close.bin,n = 300)$y
vl.big <- approx(volatil ,n = 300)$y

res <- cumsum(cl.big*vl.big)


IDEA 4

We may decompose prices and cluster volatility, i.e. reduce its degrees of freedom (for example 10 clusters (states), i.e. standardize it, and then add back the standardized volatility

 

What are you guys, like, scientists or something?

Gentlemen, you're just looking for a system to make money.

but no one, absolutely no one, is doing it.

 
Renat Akhtyamov:

What are you guys, like, scientists or something?

Gentlemen, you're just looking for a system that allows you to make money.

But no one, absolutely no one, can do it.

So maybe you just do not need to look like everyone else, the very same all who "no one, absolutely no one, it does not work.

Maybe you should stop "like everyone else", maybe you should stop optimizing these crap stochastics in testers.

Maybe it's better to watch what smart people are doing.

 
Renat Akhtyamov:

What are you guys, like, scientists or something?

Gentlemen, you're just looking for a system to make money.

but no one, absolutely no one, it does not work

Do you think only scientists can come up with something worthwhile?
Prof. Saveliev calls most scientists grant mongers)) Grants are usually valid for no more than 3 years. Something fundamental cannot be discovered in 2-3 years.
And here there are people who have been poking around the topic for a dozen years. Higher education has a majority, and it lays the foundations of scholarship)) It is just scientists, getting higher education, "got a job" in the Academy of Sciences or research institutes, rather than scientists somewhere else, but the skills have and may well in their spare time on their own something to research and invent something.
 
mytarmailS:

When scientists want to understand a complex process, they try to decompose it into simpler components and analyze them, for which spectral analysis was created.

The ideas are interesting. In my opinion the most unpredictable thing is the direction of price movement. If the results are encouraging, give me a hint, maybe others will start picking this direction.

 
mytarmailS:

So maybe just do not just look like everyone else, the very same everyone who "no one, absolutely no one, it does not work"?

Maybe it's enough "like everyone else", maybe we should stop optimizing these crap stochastics in testers.

Maybe it's better to watch what smart people are doing?

The smart ones don't leak

 
elibrarius:

The ideas are interesting. In my opinion the most unpredictable thing is the direction of price movement. If the results are encouraging, give me a hint, maybe others will start to pick in this direction.

The point is just that there are a lot of ideas, I am one, I just wrote this post in the hope that someone will be "hooked" and we have a kind of together will be

 
mytarmailS:

The point is just that there are a lot of ideas, I am alone, and I just wrote this post in the hope that someone will be "hooked" and we will sort of be together.

I was "hooked," but not by "scientificity," but by the lack of meaning. sorry. Of course, if by scientificity you mean some kind of thought process that describes the object of study in special terminology, then yes - it is a "scientific" view. However, it has nothing to do with the market, except the concepts of price and volatility, which, as a "horse in a vacuum" is considered by itself.
You have everything by itself: the price by itself, volatility by itself and all other parameters, too, by itself. It seems that the process of price formation is as random and meaningless to you as the flight of a fly. Imagine that it is not people but a fly that form the chart in two-dimensional space. To you, no difference, because the problem would again boil down to statistical prediction of random wandering...

Statistical prediction of a meaningless (through your and our common fault) random process, is useless not only scientifically, but from any adequate point of view.

My advice: return the original meaning to what is happening on the graph and find the right MO application points.