From theory to practice - page 458

 
Alexander_K2:

9. Without a parameter like ACF or entropy, there is nothing to do in the market. This is the law!

I already tried to find something about entropy, I studied information theory, but we don't know the alphabet, we don't know modulation... we don't know if there is any information in market quotes at all?

i tried to build the alphabet by searching for bar heights on the history, i found repeats on TFs up to M5, you can almost build the alphabet, but when i increase the TF, there is no recurrence of bar heights - i have two conclusions, either the bar height does not provide information or i lose information when i increase the TF

we can try to create the alphabet using other methods, but I doubt that there is informational entropy in the market quotes

Thereason why the alphabets may be different, but I doubt there is any informational entropy in market quotes. "They do not exist, google admin Renat's old posts in the threads about ticks, all the ticks you can find even for a fee, all are ticks from liquidity aggregators

If you do not know what ticks are in the terminal or what ticks we can find, there are ticks that only contain information about the price, there are deals with prices -last price in the terminal, what do you need to analyze? there are more questions than you can answer.

well, what i now consider as a model - a window, but not a sliding window, and for one day i made a logarithmic chart, i don't know why, but the levels are the same every day, and the breakdown of levels on the news, maybe it's random, but maybe the levels have some information from the market in quotes

 
Alexander_K2:

I've given a lot of energy to the market - I guess it's time to take stock now...

From the facts:

1. tick quotes come in as a kind of simple flow with p=0.5

2. To "see" a real distribution of tick increments, one should work with the sample volume of at least 1371 values, and they must be REAL, not pseudo-quotes.

3. this is achieved by working with a sliding window of at least 24 hours.

4. we practically have a Laplace distribution for the increments. Laplace motion! Do we know much about this process? No - research has just begun...

5. Sum of NE with Laplace distribution - well, actually a normal distribution with caveats.

6. It turns out that Warlock is right - we need to work with the sum of the increments in the sliding window.

7. But! There is still non-stationarity in the process. All the same, when you exit the variance channel, you need to look at some parameter - like ACF - whether or not you can enter the trade.

8. Why ACF and not entropy? Don't know - just researching...

9. There is nothing to do in the market without a parameter such as ACF or entropy. This is the law!

From the wiki, so as not to make a mess.

If the initial function is strictlyperiodic, the autocorrelation function graph will also have a strictly periodic function. Thus, we can judge from the plot about periodicity of the initial function and, therefore, about its frequency characteristics. The autocorrelation function is used to analyse complexoscillations, e.g.the electroencephalogram of a person.

Where do you find strict periodicity in markets?

 
Alexander_K2:

Seriously - I don't understand, if we don't share our experience and don't organize into teams - then why communicate on the forum?

Please understand this question - I am not the weakest physicist on planet Earth, but honestly - I am not smart enough to solve this problem. You need allies and helpers. If there are none, all is vanity and nonsense. A man can't beat the market alone...

Well, I have, so he can. To solve the problem, we have to forget physics and start studying the market.

Do you think physics is the right tool to study music, literature, psychology, medicine?

Similarly, price behaviour and particle behaviour are completely different processes.

The instrument has to be fit for the task. And you are hammering nails with a microscope. And without the slightest understanding of your own actions.

 
Alexander_K2:

I've given a lot of energy to the market - I guess it's time to take stock now...

From the facts:

1. tick quotes come in as a kind of simple flow with p=0.5

2. To "see" a real distribution of tick increments, one should work with the sample volume of at least 1371 values, and they must be REAL, not pseudo-quotes.

3. this is achieved by working with a sliding window of at least 24 hours.

4. we practically have a Laplace distribution for the increments. Laplace motion! Do we know much about this process? No - research has just begun...

5. Sum of NE with Laplace distribution - well, actually a normal distribution with caveats.

6. It turns out that Warlock is right - we need to work with the sum of the increments in the sliding window.

7. But! There is still non-stationarity in the process. All the same, when you exit the variance channel, you need to look at some parameter - like ACF - whether or not you can enter the trade.

8. Why ACF and not entropy? Don't know - just researching...

9. There is nothing to do in the market without a parameter such as ACF or entropy. This is the law!

Alexander, be careful on the corners. Where is the proof of these assertions? Or are we going to throw our caps at the market?

 
secret:

Well, I've beaten it, so it can. To solve the problem, you have to forget about physics, and start studying the market.

Do you think physics is the right tool to study music, literature, psychology, medicine?

Similarly, price behaviour and particle behaviour are completely different processes.

The instrument has to be fit for the task. And you are hammering nails with a microscope. And without the slightest understanding of your own actions.

+100. But don't be discouraged, Alexander, with your persistence it is market research that can lead to success. I am sure you can do it.

 
Igor Makanu:

Well the process you are investigating is not exactly of a physical nature, I read this forum a lot, for example you want to investigate market quotes as an electrical (physical) signal, without analysing the nature of the signal itself, you often write about distribution and sliding window... imho not there, I have loved physics since school, but precisely because it is a science of the physical world - i.e. it is tied to a physical process and each process is considered individually, another issue is that many physical models are repetitive

The market, in fact, has a physical nature, and the apparatus of physics and mathematics is quite applicable to the market. As, by the way, so is everything else.

One could talk for a long time about black boxes, their reactions to influences, distributions, statistics, etc. And it was even gathered... but it's really very long. And it all works quite realistically when applied to the market. And you don't have to go into too much trouble, even relatively simple models work quite well.

Of course, no model can give us 100% of predictions, but 40-60% is enough to make some profit. For the especially greedy )) - You won't become a millionaire, but enough to live on.

 
What if the observation window is not time, but the sum of the points?
 
Evgeniy Chumakov:
What if the observation window isn't time? What if it's the sum of the points?

Ahem... Something on the verge of genius... I'll have to think about it.

 
Evgeniy Chumakov:
What if the observation window is not about time, but about the amount of points ?

Eugene, it has been done for a long time, you are rediscovering America.

 

Gentlemen!!!

I don't know how to ask questions here to get an answer... Wall of silence...

Does anyone use entropy (non-entropy) in their algorithms? Does it make sense to use it?

I am researching it anyway, but it's a pity about the time. I've been looking for the grail for a year... It's not good.