Market patterns - page 34

New comment
 
Alex_Bondar:

Mmmmmm.... I don't know, I don't know... There's something wrong with the premise on which the model is built.

What, exactly, are you confused about?
 
yosuf:
What, exactly, are you confused about?

Well, here's a start:

Suppose further that during an infinitesimal time interval dt the acting force decreases by an amount dD(t) proportional to the remaining force D(t) by time t

In other words, you assume a simple negative feedback to continuous functions. Price time series are inherently discrete and the "forces" do not depend on their increments as easily, since they are not generated by simple potential fields as in physics.

If we remove the stochastic component of "external forces" and consider only feedbacks, there are definitely two forces, one proportional to the displacement, up to a certain threshold of the time the displacement is in one direction, the other opposite when this threshold is passed. Generally similar to the interatomic interaction.


You can see why. The effects of weighty fundamental forces (news, etc.) due to predictions and delayed reactions "smear", and statistical fluctuations that have passed the threshold are first picked up by the feedback loop (oops! trend!) and then slowed down by stops of those who missed the prediction and those who guessed, who overturned in the beginning. Exponential decay is hardly even approximately the right model. More like a mix of positive feedback and negative feedback with different coefficients.

 

I liked that from the spider

Here you are an economist, you were once touted as the most sensible of the alternative economists to the Americans. Can you at least explain why there is no sensible textbook on market economics at all, but all bullshit? Or maybe there is?
Andrey

About ten years ago I decided to write a good market economy textbook. I wondered why no one has written one yet. I decided, as a mathematician, to strictly follow models, not to draw crosses like all textbook authors. I honestly built a model of Bem-Bawerk's bargaining, a model of market price formation in general. I started to prove the convergence of processes, ergodicity...

And then I discovered that the processes of such a class are in the general case irreducible. That is, in the general case the market price simply does not exist. It follows that market regulation is possible only in a very limited class of sectors. In the rest, it inevitably entails imbalances and therefore must be subject to external, violent adjustment.

Then I also discovered the impossibility of forming a universal equivalent in a natural way. The price matrix was in general asymmetric. Consequently all the theories of the natural origin of money are groundless.

It is the same with the case law market models. As soon as I began to solve them analytically instead of drawing crosses, it turned out that there is no equilibrium in the markets, a generally divergent process, like Friedman's universe: without a cosmological term there is no equilibrium to be fitted. So the whole Keynesian theory generally speaking turned out to be a bluff.

And a lot of anecdotes like that came out. After which I gave up the idea of writing a textbook on market economics. For the lack of a real subject matter to describe.

Modern market views are best described in the textbook "Macroeconomics" by Gregory Mankew. Clearly, it is intended to educate clerks, and has nothing to do with real business.

 
yosuf:

Gamma ray is something else entirely. But it wouldn't matter if the indicator showed something useful, but I haven't seen anything like that yet.

 
lucky_teapot:

Gamma ray is something else entirely. But it wouldn't matter if the indicator showed something useful, but I haven't seen anything like that yet.

Here is the run on the tester (so far) from 2009 to 2013 with SL=TP. Please tell me, what allows us to get the advantage if it is not due to the Gamma distribution, which I have specifically included in the indicator at a constant lot 0.1?

And at R=0.0005:


 
Alex_Bondar:

Well, that's it from the beginning:

In other words, you are assuming a simple negative feedback, and to continuous functions. Price time series are inherently discrete and the "forces" depend, among other things, on their increments not so easily, since they are not generated by simple potential fields as in physics.

If we remove the stochastic component of "external forces" and consider only feedbacks, there are definitely two forces, one proportional to the displacement, up to a certain threshold of the time the displacement is in one direction, the other opposite when this threshold is passed. Generally similar to the interatomic interaction.


You can see why. The effects of weighty fundamental forces (news, etc.) due to predictions and delayed reactions "smear", and statistical fluctuations that have passed the threshold are first picked up by the feedback loop (oops! trend!) and then slowed down by stops of those who missed the prediction and those who guessed, who overturned in the beginning. Exponential decay is hardly even approximately the right model. More like a mix of positive feedback and negative feedback with different coefficients.

1. The continuity of the FWR is a forced assumption to make it easier to apply the mathematical apparatus. Subsequently, it became clear that this assumption is negligible with sufficient sampling.

2. It is assumed that the function D(t) can be both positive and negative and then everything converges to your reasoning about positive and negative feedback.

 
yosuf:

Here is the tester run (so far) from 2009 to 2013 with SL=TP. Please tell me, what then is the advantage, if not due to Gamma distribution, which I have included in the indicator body at constant lot 0.1?

And at R=0.0005:

So why are the signals so bad?


In my opinion, the premise is very doubtful. And the result is some kind of alchemy. Perhaps there is some common sense there, but it is clearly not in the exponents.

But I would like to play with it, if there is an indicator for MT5. I didn't find one.

 
lucky_teapot:

So why are the signals so bad?


In my opinion, the premise is very questionable. And the result is a kind of alchemy. There may be some common sense there, but it's clearly not in the exponents.

But I would like to play with it if there is an indicator for mt5. I have not found one.

1. This mode has been connected since 13. 08. 13

2. see the indicator for mt5 in the kodobase https://www.mql5.com/ru/code/10339.

 
lucky_teapot: the data from there is broken down into archives and spread out into subdirectories half an hour long on average.
I'll stick it in the Annals, it's too juicy.
 
Mathemat:
I will put it in Annals, it's too juicy.

Really funny thing to say))

yosuf:

1. this mode has been connected since 13. 08. 13

2. see the indicator for mt5 in kodobase http://codebase.mql4.com/ru/7638 .

Thanks.

I'm only a beginner, so I might say something stupid, but in my opinion indicators-extrapolators, should be implemented in two versions, one is for humans, that is some curve(s) with instructions for its (their) interpretation and the other is for machines,2d probability vector of potential displacement. X-which way, Y-which probability. Using this vector on a particular bar (tick) without unnecessary thinking the TS is built (tested) and all questions disappear about whether it catches any patterns or not.

I don't understand how to reduce it to these two numbers in this realization of your indicator, that's why I can't adequately evaluate the effectiveness of the algorithm.

Please formalize it, i.e. every bar corresponds to such a vector.

1...272829303132333435
New comment