![MQL5 - Language of trade strategies built-in the MetaTrader 5 client terminal](https://c.mql5.com/i/registerlandings/logo-2.png)
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
On what grounds are you dissatisfied with linear regression? After all, there have long been methods of justifying the need to increase the complexity of the functional form of the model, always checking whether the chosen functional form is redundant.
Now you're piling up the time, and now you're throwing out words about linear regression ...... More helpful, pls.
You are the one dealing with time series. And I have the same question again - where is the justification that price = time series? Because it's convenient for you? Because that's the way it's usually done?
By definition. A time series is a sequence of observations ordered in time.
Price, when viewed in terms of time, as is usually done, is a martingale ( https://www.mql5.com/ru/articles/1446 ). Or very similar to it. As such, no reasonable strategy allows you to "profit" from it.
If price is viewed from the perspective of the processes that form it, it is not a martingale.
From the point of view of many models of liquidity provision - the price is really a martingale in relation to some information sets (for example, the flow of transactions on the corresponding instrument). At the same time, that there are inefficiencies in markets that can be exploited. And this means that there are information sets, relative to which the price may not be a martingale.
Below I give an example of how one of the quoting models works. These prices are martingale with respect to absolutely all previous prices and flow of deals that were performed. The deals flow is generated randomly. I.e. it is impossible to take money away from such a quoting algorithm. At the same time, if several instruments are quoted simultaneously or if the stream of deals is not random - these processes will not be martingales. In this case relevant factors can be taken into account in the model, and again prices of all instruments will become martingales relative to the available information set.
You are the one dealing with time series. And I have the same question again - where is the justification that price = time series? Because it's convenient for you? Because that's the way it's usually done?
Hover your cursor over any bar and you see the word Time at the top of the table. You open almost any indicator - there is the word Period or its analogue. Or is it a practical joke for you?
If the price is considered in terms of its forming processes - it is not martingale.
I cannot think of any function (model) that has as its argument a process that forms the price of any asset, not to mention currencies.
The linear relationship does not fit the logic of the process.
If we know the logic of the process. and if this is known, there is no problem at all.
I am proceeding from considerations of model sufficiency and redundancy. If we use AR(1) model, i.e. I make a prediction on the value of the previous bar, then what does it have to do with any complexity above linear? If my model considers 100 bars, it cannot be approximated by a straight line. If you build the model from the second Nicolachean, you have to strain it too much.
So? What successes do these sound methods have?
Nothing follows from the fact that you don't know them and don't seem to be able to understand them.
Yusuf and Ananimus and I have touched on a couple of related topics. They have nothing to do with you.
Pardon me, although I seem to be the topicstarter.
Once again:
Mathemat: There is no need to confuse root cause and dumb independent variable.
Do you want to drive or do you want to drive?!
You are the one dealing with time series. And I have the same question again - where is the justification that price = time series? Because it's convenient for you? Because that's the way it's usually done?
Price, when viewed in terms of time, as is usually done, is a martingale ( https://www.mql5.com/ru/articles/1446 ). Or very similar to it. As such, no reasonable strategy allows you to "make money" on it.
If price is viewed from the perspective of the processes that form it, it is not a martingale.
There you go again with your reasoning... The same was in the feature selection thread, only there you had to justify the applicability of TI to the market.
"Price = time series" simply because it's convenient has been around for at least 500 years. No mathematician or practitioner gets into this medieval philosophical nonsense and tries to claim that in a process described by an explicit function of time f(t), time is the prime cause.
So one can go so far as to say that in the motion of a stone thrown upwards, time is the prime cause. But this is not true: Newton's laws do not explicitly contain time, they are the relations of cause and effect. However, from these laws many divergences are derived in which the independent variable is time. This is convenient - full stop.
Enough with the sophistry which is of no use. There is no need to confuse root cause and dumb independent variable.
Martingale or no martingale is simply a way of looking at it or a matter of faith. There are several schools of thought that view price in different ways. The first is fundamentalist, they look at formative processes and believe they know the causes. They often don't even know what a martingale is, as they don't need it and it is even harmful.
The second are chartists (thechanists) who abstract away from these alleged causes and stupidly look at the price as a function of time (or bar number, for example) and try to predict it. It's convenient for them, that's all. They don't care at all whether time is the root cause of price movement or not - just like fundamentalists, by the way.
There are other schools, but let's not talk about them.
To what you've written, Alexey, I want to add the objective as a justification for using the price.
When using any econometric package, the first problem is the following: take quotes from the source as a sequence of prices, or label these prices with time stamps? If not labelled, it is easier. Monday immediately follows Saturday and the daily price starts at 00:01 minutes. If you go with time, then all these nuances will have to be sorted out, and who will pay for this extra effort, which by the way is not easy? And the rationale is in the model used: if the model is of TA type, time stamps are not needed, but if we want to take into account our information about asset price cyclicities, for example sales of c/o products? After all, this is important enough information to ignore. In this case, the model explicitly includes time as a variable.
Maybe HideYourRichess has never built models with cyclicality or other direct use of time as a function argument? This is not surprising for TA.
faa1947:
It's no surprise to TAs.
To what you have written, Alexei, I would like to add the objective as a justification for using prices.
When using any econometric package the first problem is the following: take quotes from the source as a sequence of prices, or label those prices with time stamps? If not tagged, it is easier. Monday immediately follows Saturday and the daily price starts at 00:01 minutes. If you go with time, then all these nuances will have to be sorted out, and who will pay for this extra effort, which by the way is not easy? And the rationale is in the model used: if the model is of TA type, time stamps are not needed, but if we want to take into account our information about asset price cyclicities, for example sales of c/o products? After all, this is important enough information to ignore. In this case, the model explicitly includes time as a variable.
Maybe HideYourRichess has never built models with cyclicality or other direct use of time as a function argument? This is not surprising for TA.