Market etiquette or good manners in a minefield - page 68
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
OK, apparently this work is so worthy of my attention that the wait could go on forever. :) Anyway, I suggest you find and read Bezruchko, Smirnov - Mathematical Modelling and Chaotic Time Series. If you haven't read it, of course. These are pros who write, easy to read, easy to understand, accessible to all:)
ok. Got it. Thank you.
The Russian mathematician Euler, my idol, wrote about 800 papers, each of which is as important and novel as a modern dissertation.
Mathemat , why don't you come down?
You have a problem to solve - to amuse yourself!
ok. Got it. Thank you.
Report the result of familiarization with the secret material!
Let's discuss.
As for learning, I'm sure that thanks to these conversations with you and the forum members involved, I'm learning myself and the process is endless... I hope.
>> Thank you!
I'll work on the representation of the profitability function from the input dimension.
P.S. I would like a popular introduction to Pastukhov's development, because without mathematical education there is nothing to do - it's like reading hieroglyphs, but I want to proceed to more real experiments on adequate BP. Maybe you don't want to discuss it - then be on your way.
to Mathemat
Really, Alexey... can you do it? The crowd is asking... -:)
Of course, you have to respect the community.
What is the task? I look at this thread with my left eye, sometimes inserting my smart comments, but I do not get into the essence - only to discourage not too smart colleagues who have cut in here with overly profound questions.
I'll finish my congenial advice, put it on the demo, and then see, okay? It'll take me a couple of days.
Let's wait! We have plenty of time.
In the meantime, for the subconscious, the condition of the problem:
There is a black box with d tentacles sticking out into the street, with which it learns the world. The box has a total of w configurable parameters (with w including d). Question: How many parts of some object does the box need to feel to identify the object as plausibly as possible?
The answer such as "the more, the better" does not work, because everything the box learns about the object, it stores in the same parameters and they simply may not be enough for everything. The answer - "the less, the better" - also does not work, because the minimum of information is its absence. Consequently, it is impossible to construct a plausible judgment about the subject in this case. This means that there must be a golden mean (optimum) in the number of attributes. So what is it equal to?
P.S. I think I began to understand the equality P=w*w/d see fig:
You can see, if this equality is fulfilled, prediction variance decreases and prediction accuracy (blue line minimum) doesn't worsen yet (in the next step, which is not presented, the grid is not so well trained). In fact, the variance directly determines the risk of loss - the greater the range of the forecast result, the greater the risk relative to the deposit we take, and therefore the less leverage we have to use, which in the end reduces the profitability of trading in the MTS+MM combination.
The obtained result is not transparent and the question remains open, as I have formulated it.
In general, the task of trading, in my opinion, boils down to solving three main problems:
1. rejection of the Efficient Market Hypothesis. - Do I understand you correctly that rejection of the EMH hypothesis, automatically leads to quasi-stationarity of market BPs?
2. Understanding that of the three ways to make money:
...
c) analyzing historical instrument price data,
the last two are available to us. Of them, the analysis of macroeconomic news does not imply the use of MTS (at least I do not see a way of real implementation). So that leaves the last point.
Here I agree with you, indeed the network is similar to a non-linear AR model, with one exception (as you rightly pointed out) - it does not require a priori knowledge of the process model, and this is its biggest advantage given the non-stationarity of market BPs. Of course, we mean their quasi-stationarity (according to the first point), otherwise no NS can work. Thus, I don't care whether or not the condition is met: "the current price is a non-linear function of past prices", in general the non-linear NS will allocate the right area. Hence, there are no alternatives for NS!
More precisely, it does not follow from a statement of quasi-insteadiness that the market is not fully efficient.