Machine learning in trading: theory, models, practice and algo-trading - page 2677

[mytarmailS]

Maxim Dmitrievsky #:
There's one, loess regression, I think. It does it better. Detrending should leave maximum information, I already wrote above

There's a lot of things. It's no use.

[Valeriy Yastremskiy]

Aleksey Nikolayev #:

The analogy is quite understandable, but it is not fully developed) Each "particle" of the market reflexes, tries to comprehend the market as a whole, etc. (like your reasoning, for example). This changes everything a lot and it is hardly possible to "catch" this "physics" by simple wave approaches.

Every cricket (particle) should know its own place). Of course, after the assumption that there are layers of identical traders, it would be good to have data from the exchange or all exchanges)))) on each tick of transaction slices by amounts / quantity and by types of orders and life would be easier. But this is indirectly criminal information to disclosure, although I saw revelations in the Internet on the initial actions to analyse this information in the sum over the years and the number / volume of transactions for the formation of algorithms TS).

Of course, simple wave trades will not catch. But we have approximately the same thoughts, to single out a certain long-playing low-omplitude signal / component or to build a picture of signals' behaviour in dynamics taking into account their recognition on slices and detection of fades, appearance of new ones and finding long-playing ones.

[vladavd]

Maxim Dmitrievsky #:
Usually they are found on a hunch and do not live long )

So, since the topic is about the MoD, I like all sorts of search of options, sometimes it's not bad to find one

A signal is always found in the same way (almost), but its life is very different: it can work perfectly and for a long time (if it is received on a large TF), but it can also turn out to be a fake, and it can also change significantly along the way. So there is no honey anywhere, unfortunately).

All in all, these are just ideologically different approaches. But I personally find the search for inefficiencies less reliable, because it is not the inefficiencies that are primary, but the signal around which they arise. Inefficiencies are like small predators that come to eat the remains of the prey that the lion has taken. They make the weather a bit, but without the lion they wouldn't be here at all. Though, of course, a pack of hyenas can sometimes chase away a lion, but the opposite happens much more often.

[vladavd]

Alexsey Minchenko #:

The goods are purchased at certain price levels, for this purpose time series (price changes at suppliers) are analysed, then we buy from the one who is cheaper and then sell to the one who has a higher price. And there is no need to keep a warehouse 😄. I see it like this ))

Well that's a pretty narrow special case; for someone to barter, someone else has to do real things in the real world. This spread will converge one day (the inefficiency will close) and the free profit will evaporate, you will have to look for a new one, and whether you will be able to find one is a question, because there are few fools and nobody wants to give their money for nothing. And the manufacturer will continue to do it as it worked.

[sibirqk]

Price changes, in my opinion, are the results of deals made in one or another direction. There are many multidirectional transactions in the market at the same time in a unit of time. The best model to describe this, in my opinion, is random walk. But probabilities to describe such a wandering are not stationary, they themselves change in time. Actually, these probabilities only have a physical meaning when describing the market. And the price chart that traders look at is integral in relation to time dependences of probabilities. It would seem that differentiate the price chart, i.e. analyse returns and voila, you will be happy. But everything is not so simple. The current probability is the sum of probabilities of transactions made by different groups of market participants. In principle, there are not too many such groups - long-term investors - all kinds of funds that focus on the interest rate differential, banks, industrialists, medium-term speculators, short-term speculators and so on. Trading strategies in each group are not very different and the probabilities of making trades within the groups are linear and stationary for some time, although sometimes they can switch quickly. In principle, in my opinion, Kalman is ideal for predicting the probabilities of making trades in each of the groups of market participants. The problem here is how to separate them, i.e. how to extract the statistical characteristics of each group. In principle, one of the most effective approaches, in my opinion, is to use Kalman in conjunction with machine learning methods, but I don't really know how to do it.

[mytarmailS]

sibirqk interest rate differential, banks, industrialists, medium-term speculators, short-term speculators and so on. Trading strategies in each group are not very different and the probabilities of making trades within the groups are linear and stationary for some time, although sometimes they can switch quickly. In principle, in my opinion, Kalman is ideal for predicting the probabilities of making trades in each of the groups of market participants. The problem here is how to separate them, i.e. how to extract the statistical characteristics of each group. In principle, one of the most effective approaches, in my opinion, is to use Kalman in conjunction with machine learning methods, but I don't really know how to do it.

#26694

velocity separation (hft, mid-range, long-range).

Amplitude separation.

[Maxim Dmitrievsky]

vladavd #:

A signal is always searched for in the same way (almost), but its life can be very different: it can work perfectly and for a long time (if it is received on a large TF), but it can also turn out to be a fake, and it can also change significantly along the way. So there is no honey anywhere, unfortunately)

All in all, these are just ideologically different approaches. But I personally find the search for inefficiencies less reliable, because it is not the inefficiencies that are primary, but the signal around which they arise. Inefficiencies are like small predators that come to eat the remains of the prey that the lion has taken. They make the weather a bit, but without the lion they wouldn't be here at all. Although, of course, a pack of hyenas can sometimes chase away a lion, but the opposite happens much more often.

There can't be a signal on a quote chart, it would mean that there is information about the future contained in it. But since the chart is just a reflection of the actual exchange rate in the moment and for past periods, it does not contain any information about the future. Inefficiencies do not occur on top of a signal, they are situations that have a high probability of repeating with the same outcome several times or many times. They are not a signal either, as they only reflect some other hidden or explicit processes.

Inefficiencies are information undiscounted by the market.

The notion of signal is superfluous here, because it does not lead to simplification of understanding and does not add new information. But the concept of an efficient market fits very well, because it is informative and closer to the topic, and it is broader.

[mytarmailS]

Maxim Dmitrievsky #:
There can be no signal on a quote chart, which would mean that there is information about the future that it contains. But since the chart is just a reflection of the actual exchange rate at the moment and for the past periods, it does not contain any information about the future. Inefficiencies do not occur on top of a signal, they are situations that have a high probability of repeating with the same outcome several times or many times. They are not a signal either, as they only reflect some other hidden or explicit processes.

The notion of a signal is superfluous here, because it does not lead to simplification of understanding and does not add new information. But the concept of an efficient market fits very well, because it is informative and closer to the topic.

)))) you have overdone something with definitions

[Maxim Dmitrievsky]

mytarmailS #:

)))) you're overthinking the definitions.

It's economics, son. Blue diploma 🤣

Elasticity of supply and demand

[mytarmailS]

Maxim Dmitrievsky #:
It's the economy, son. Blue Diploma 🤣

Elasticity of supply and demand

I've got it, question answered, there is no signal in the market.

)))))