Machine learning in trading: theory, models, practice and algo-trading - page 1767

 
Valeriy Yastremskiy:

In the folder DOC 7zFormat.txt look. It's complicated.Maxim Dmitrievsky help, if not lazy))))

What is it? In my article I have a code for determining the entropy of a series and a link to the article
 
Maxim Dmitrievsky:
What is it? I have a code to determine the entropy of a series and a link to the article
No, this is high matter, here below. From the archive file you need to take out the compressed sections and uncompress them back with a time reference.
 
Valeriy Yastremskiy:
No, this is high matter, it is lower. You have to extract the compressed portions of the archive file and unarchive them back with time reference.

So it will take out random chunks where entropy is lower... and what's the point of archiving?

the purpose is not clear
 
Maxim Dmitrievsky:

Well, it will take out random pieces where entropy is lower... so what's the point of archiving?

purpose is not clear
To refute the theory that the archiver can determine the patterns. Or to confirm. Although you're right, if we pull flat trends, we will see that the archiver defines them, but what will it give us in terms of forecasts.... The archiver works on history))))
 
Mihail Marchukajtes:
If you do not know the basic data it is required. Those data that are transmitted by the exchange. All without exception and important .... I understand that this tool writes history on its own or can pull it out of the copycat?

Yes writes (wrote) the data on the OI (open session volumes, the glass), which are not saved. And there are ticks in the history, there is no need to write them, any indicator can draw all these deals. I also imposed the balance of deals directly on the price, it also worked out very well (divergence). If I look at the chart I will see the difference between the two indicators.

 

I plan to return to NS, while I am still learning. The basic principle of my TS will most likely be based on the main law of financial markets. What is it? " The finmarket gets rid of patterns, trends. For this reason it is very difficult to predict, because the bulk of searchers/researchers are trying to find patterns statistically (then extrapolating), or just playing in a foreseeable floating time window. On the second: Cotier draws out the visual shapes we see with our eyes. If in the near future we see the beginning of a repetition of these figures (uniformity, "regularity"), we subconsciously expect this "trend" to continue. These can be figures that our bio-NS has memorized (all sorts of heads, shoulders, W, M, trends, pullbacks, whatever - succinctly "patterns"). That is, my main idea: the sign of hooking is the proximity of visually similar patterns, where the beginning of the subsequent "same pattern" is the hook. If there is a good "nibble," the crappie moves against the poor fish. That is, the pattern can continue to be drawn correctly if there is no real volume (fish). Look at these drawn charts, in places this catch is clearly visible...

The principle of a smart TS should probably take this pattern into account. To determine the most frequently recurring (and memorable to us) shapes (patterns) the SOM neural network is suitable, because it can classify them independently. Further - work with a floating chart window, classification of arriving "pictures", their recurrence, estimation of probability of biting.

In general, we get out of the water and take a fishing rod in our hands :) These are the ideas. I will work in this direction...

 
MrBobr1:
There is such humor. A warrant officer asks a soldier. What is the difference between the moon and the sun. The soldier begins to explain that the sun is a star and so on. The ensign interrupts him with the word "fuck" and says that the sun shines in the daytime and the moon at night. We, as traders, are not interested in all the regularities, but only in those that we can use for making money. Here is a regularity at important news, there will be a strong price movement. But this law is of no use to us. There will be a movement, but we don't know in what direction. Sometimes the price goes not only against the news, but against common sense. But I, like many traders, use this pattern, though I don't want to lose money. I step out of positions or do not open before the news. I can't find it, put down a case of good whiskey, I will help you.
You want to know a secret? Only you don't tell anybody. Deal. There are market-neutral strategies that do not care where the price goes, as long as it does. Why not use them in your example?
 
Sealdo Sergey:

Yes writes (wrote) the data on the OI (open session volumes, the glass), which are not saved. And there are ticks in the history, there is no need to write them, any indicator can draw all these deals. I also imposed the balance of deals directly on the price, it also worked out very well (divergence). If I look at the chart I will see the difference between the two indicators.

I am very interested in new data connected with Moex. But to write OM in a file is a dead end, as myself have such a note, but in the consequence that something to use has not turned out. In the rest is ready to develop this topic, as well as actually specified by you and are fundamental to the future price changes. I also need to add a smile, and it seems that it can be pulled from Quicksilver, but it is all programming, which I am not very good at. I need to look for a team :-( But this should go to the forum of programmers. Apparently there are no programmers here :-(
 
Sealdo Sergey:

I plan to return to NS, while I am still learning. The basic principle of my TS will most likely be based on the main law of financial markets. What is it? " The finmarket gets rid of patterns, trends. For this reason it is very difficult to predict, because the bulk of searchers/researchers are trying to find patterns statistically (then extrapolating), or just playing in a foreseeable floating time window. On the second: Cotier draws out the visual shapes we see with our eyes. If in the near future we see the beginning of a repetition of these figures (uniformity, "regularity"), we subconsciously expect this "trend" to continue. These can be figures that our bio-NS has memorized (all sorts of heads, shoulders, W, M, trends, pullbacks, whatever - succinctly "patterns"). That is, my main idea: the sign of hooking is the proximity of visually similar patterns, where the beginning of the subsequent "same pattern" is the hook. If there is a good "nibble," the crappie moves against the poor fish. That is, the pattern can continue to be drawn correctly if there is no real volume (fish). Look at these drawn charts, in places this catch is clearly visible...

The principle of a smart TS should probably take this pattern into account. To determine the most frequently recurring (and memorable to us) shapes (patterns) the SOM neural network is suitable, because it can classify them independently. Further - work with a floating chart window, classification of arriving "pictures", their recurrence, estimation of probability of biting.

In general, we get out of the water and take a fishing rod in our hands :) These are the ideas. I will work in this direction...

I have a fully working data preparation technology and it works on the data I use. Delta, volume, price. So I suggest we join forces and expand the scope of the data I get from the market, thereby improving an already well working system. How about you?
 
Valeriy Yastremskiy:
To refute the theory that the archiver can determine patterns. Or to confirm. Although you're right, we will pull flat trends, we will see that the archiver determines them, but what will it give us in terms of predictions.... The archiver works on history))))

Ehhhh and how well it started)))) entropy, regularities)))) regularities only not those turned out, all right, we compress only the same color, the same frequency, the same increments)))) And as a fact, first the archiver finds an area with the same values and only then compresses them. Shit.... And we're looking for the wrong patterns. I read Panchin's today.)) With extraordinary ease we confuse meaninglessness and profundity))))))