1st and 2nd derivatives of the MACD - page 29

New comment
 
Farnsworth: then the 'green', inexperienced, but unrelenting can be called - quantus :o)
That's Yusuf for sure. Quantus is number one on the forum.
 

Mathemat:

Are you sure you got it right?

Doesn't match the Wiki, but the meaning is correct. General algebra refers to axioms, which are not to be proved by means of the theory itself.

That, to put it mildly, is also an exercise in verbal equilibrium. Nobody prevents to consider minutes or even ticks and draw conclusions from them that contribute to a profitable strategy. That's what quants do.

The article draws conclusions on a minute interval. You can use ticks, but conclusions on longer intervals

 
faa1947:

Doesn't match the Wiki, but the meaning is correct. General algebra refers to axioms that are not to be proved by means of the theory itself.

And wrong in meaning. Reread it again and see if the blue highlighting is in place:

I would like to remind that any theory is inconsistent only if it contains propositions which cannot be proved by means of the theory itself.

 
Farnsworth:

A quote is a complex stochastic multifractal, it is not even self-similar, but on a trader's accessible "visual" scale it behaves almost like a martingale, even though the process has strong non-linear relations. One can get some adequate knowledge about the process only with the help of fractal analysis, there are already many techniques and methods accumulated. For example, a singularity spectrum that will show the whole hell of the situation.)


I see 4 ways of creating trading systems

  1. To string some sense from one branch of science (economics, management systems, etc) onto currency prices and trade with appropriate methods from that branch (econometric, Kalman filters, etc).
  2. Try to mathematically understand the essence of price series by applying statistics, chaos theory, multifractal, etc. and trade with appropriate methods. Examples, please?
  3. Stop searching for the essence of the price process and simply create a neural network with certain inputs.
  4. Or even easier to do nothing special and track the price movement with oscillators like MAKD or stochastic and trade by the conventional TA methods.

I have such experience. The more complex the trading system, the quicker it will fail. You need something very simple and it should work reliably. Thoughts?

 
Mathemat:

And the meaning is wrong. Re-read it again and see if the blue highlighting is in place:

It seems correct to me. There are many formulations of the theorem in the literature. For example:

Any system of mathematical axioms from a certain level of complexity is either internally inconsistent or incomplete

The one I cited may not be mathematically accurate, but it captures the essence: in any theory something must not be proved. If everything is proved, then it is contradictory.

I am too lazy to look up the formulation in general algebra, which I think is the most accurate, because it deals with the principles of all theories.

In our activity, my formulations are quite sufficient for me.

If it is principled for you, I am ready to discuss your formulation. I do not like Wiki's formulation.

 
gpwr:


I see 4 directions for creating trading systems:

  1. To stretch some sense from one branch of science (economics, management systems, etc.) onto currency prices and trade with appropriate methods from that branch (econometric, Kalman filters, etc.).
  2. Try to mathematically understand the essence of price series by applying statistics, chaos theory, multifractal, etc. and trade with appropriate methods. Examples, please?
  3. Stop searching for the essence of the price process and simply create a neural network with certain inputs.
  4. Or even easier to do nothing special and track the price movement with oscillators like MAKD or stochastic and trade by the conventional TA methods.

I have such experience. The more complex the trading system, the quicker it will fail. You need something very simple and it should work reliably. Thoughts?

1. Revise mathematical models using economic theories - more accurately, fundamental analysis.

2. Apply any complex mathematical methods at the stage of TS creation. It is desirable to work in Matlab, so that there are no limitations on the methods.

3) I do not believe in NS - the black box.

4, At point 2 to create TS with a minimum of parameters at the expense of the theory ensuring its stability

 
Farnsworth:

PS; briefly, wrote faa: ... the quote is a complex stochastic multifractal, it's not even self-similar, but at the trader's accessible "visual" scale it behaves almost like a martingale, and this despite the fact that the process has strong non-linear relationships.


Shocker. This should immediately be written in the annals as an example of abstruse and incomprehensible verbiage. In fact, a quote is the exchange rate of the currency against the counter-currency at a certain time.

Keep it simple.

 
faa1947:

1. Cross-check the mathematical models with economic theories - more accurately with fundamental analysis.

2. Apply any complex mathematical methods at the stage of TC creation. It is desirable to work in Matlab so that there are no limitations on methods.

3. do not believe in NS - black box

4, At point 2, create a TS with a minimum of parameters at the expense of the theory, ensuring its stability.

This is where we coincide. Well, almost the same. It's not a matter of belief/disbelief either :-))

For those who like to build NS, you have to answer a few philosophical questions to begin with. Depending on the answer to it, it will be clear whether it is necessary to build NS.

NS is an attempt to repeat in a very simplified form the self. What makes the author build his pathetic semblance and not use his brain?

Any excuses for this question, like iron has no doubts and reflexions and better memory, fall on the next question. Why does an author with such a pathetic brain with doubts and reflexions and other problems that prevent him from living, trading and working think he can create another brain that is better. If he cannot cope with his own head, how can he create something better? And, if his head is fine, why build an NS?

Maybe it's simpler than that?

==================

The good posts have disappeared. Why...? :-(

 

Хорошие посты исчезли. Зачем?... :-(

It's good that at least people don't disappear...

One technical post is being restored. The rest are flaming or repetition of what has been said before.

Zhunko 11.01.2012 18:36

gpwr:


I see 4 directions of trading systems creation:

  1. To draw some sense from one of the branches of science (economics, control systems, etc.) on the currency prices and trade with the corresponding methods of this branch (econometric, Kalman filters, etc.).
  2. Try to mathematically understand the essence of price series by applying statistics, chaos theory, multifractal, etc. and trade with appropriate methods. Examples, please?
  3. Stop searching for the essence of the price process and simply create a neural network with certain inputs.
  4. It is even easier not to make any widdles at all, and track the price movement with some oscillators as MACD or stochastic and trade by the generally accepted TA methods.

I have such experience. The more complex the trading system, the quicker it will fail. You need something very simple and it should work reliably. Thoughts?

I posted a picturehere. It shows the inputs on the 2nd derivative. Filters are indices.

We will test it on real account in a month. It works rather good. At present time PF = 3. We are still tuning parameters. There is a potential to increase TF to 10.

 
Zhunko:

For lovers of building NS, a few philosophical questions must first be answered. Depending on the answer, it will be clear whether it is necessary to build NS.

An NS is an attempt to replicate, in a very simplified form, itself. What makes the author build his pathetic semblance and not use his brain?

Not really. According to Kolmogorov's theorem any continuous function of many variables can be represented as a sum of non-linear transformations of continuous functions of one variable. In other words, any continuous function of many variables can be represented by a neural network with two hidden layers h and g as shown below

Someone later derived the theorem that a NS with three hidden layers can model any continuous function. Properly chosen continuous functions are any non-linear smooth functions, e.g. h and g in most cases tanh, although simple exp can also be used.

NS are not used to model the brain, but to model the object, in our case the market. Their strength is in their versatility. It means that it's not necessary to know the laws of price movement, write diffusions or make up different regression models like (18). Theoretically, NS is able to simulate any non-linear dynamic system. In practice people often mistake the strength of NS thinking they can add prices or different indicators to network inputs and the network will find all regularities and generate trading signals. In fact, we should choose the network inputs correctly. For example, if we believe in support and resistance levels, then inputs should be based on those predefined levels and past prices in relation to those levels. The network itself isn't able to detect the importance of these levels, it should always "prompt" in which price transformation space we are looking for our non-linear market model.

1...222324252627282930313233343536...67
New comment