Machine learning in trading: theory, models, practice and algo-trading - page 2777

 
The clown is clowning around, they don't have time to ban. One goes away, another comes back, and so on and so forth.
I don't think there's any point in going on
 
Maxim Dmitrievsky #:
The clown is clowning around, they don't have time to ban. One goes away, another comes back, and so on and so forth.
I don't think there's any point in going on

Don't get distracted.

Do your science work.

 
Uladzimir Izerski #:

If you are sure, it doesn't matter what you trade on.

I'm saying under the shf, you shouldn't be trading in real money..

Every time I write, you're talking about something else.

To my first question about the marks, and to my replica.
 
Evgeni Gavrilovi #:

Could you publish any such predictor?

Maybe it is the difference in RSI indicator reading on EURUSD and GBPUSD?

The same RSI, but with some nuances.

Once again: it is not the list of predictors that is important, but the algorithm of selecting these predictors. When the window is moving, the mentioned RSI may or may not be used. This is the problem, and the first of a whole set of problems.

 
Uladzimir Izerski #:

Don't get distracted.

Do your science work.

Scientific work is welcome, but indirect advertising of your extraordinary indicators from the market is banned. Not the indicators are banned, but the author of paid indicators.

You should be more modest.

You have been on the market for six years, and ten years ago most forum members were convinced that it is extremely difficult to create a trading system working for more than 6 months within the framework of technical analysis. If we judge by the life of signals, some people out of thousands of signals succeeded, but the result is always the same - a plum of deposit. You have not provided evidence of the possibility of using your indicators, at least at the level of a state, or better, a signal.


We are engaged in MO because of practical and, most importantly, theoretical impossibility of using technical analysis. It is not from a good life that we started to get into MO. MO is a mountain of sophisticated mathematics and software. There is no love for science here.


PS.

The posted indicator is not the first one.

The signal is on the second candle, the position will be opened at the close of the 3rd candle. Profit only in the case of trend continuation above 3 candles, at least on the fourth candle.

The reversal signal will appear on the second candle, the deal will be closed on the third candle. The indicator is nothing.

Positive increments of one sign over 3 candles in a row is extremely rare, you can easily get the relevant statistics. The corresponding statistics is called autocorrelation. Usually less than three.


 

СанСаныч Фоменко #:

The relevant statistic is called autocorrelation. Usually less than three.

I read in a book from the 70s that it is impossible to predict a series without autocorrelation. Is there anything more modern on this subject?

 
СанСаныч Фоменко #:

Scientific work is welcome, but indirect advertising of your extraordinary indicators from the market is banned. Not indicators are banned, but the author of paid indicators.

You should be more modest.

You have been on the market for six years, and ten years ago most forum members were convinced that it is extremely difficult to create a trading system that works for more than 6 months within the framework of technical analysis. Judging by the life of signals, some people out of thousands of signals managed to do it, but the result is always the same - a plum of deposit. You have not provided any evidence of the possibility of using your indicators, at least at the level of a state, or better, a signal.


We are engaged in MO because of practical and, most importantly, theoretical impossibility of using technical analysis. It is not from a good life that we started to get into MO. MO is a mountain of sophisticated mathematics and software. There is no love for science here.


PS.

The posted indicator is not the first.

The signal is on the second candle, the position will be opened at the close of the 3rd candle. Profit only in the case of trend continuation above 3 candles, at least on the fourth candle.

The reversal signal will appear on the second candle, the deal will be closed on the third candle. The indicator is nothing.

Positive increments of one sign over 3 candles in a row is extremely rare, you can easily get the relevant statistics. The corresponding statistics is called autocorrelation. Usually less than three.

The properties of fractals allow to get longer correlations, for the time of existence of patterns. But you have to adjust the windows. In essence, the method consists in selecting such states by dividing the series into 2 parts and subtracting one from the other in reverse order, reversing chronologically the second or the first part, i.e. mirroring the graph at the point (attractor). These phenomena are what manifest themselves in Fat Tails and memory. That is, the pieces of the graph to the right and left of the attractor are highly correlated.

The process is formed simply: 1. Bifurcation point, past pattern breaks, new influencing information enters the market. 2. After the information is partially already accounted for by the market, an attractor is seen to forecast the remainder. 3. Collapse again, the previous dependencies stop working. The beginning of a new bifurcation is also partially predicted.

There is no such thing in classical econometrics. But it is possible to use its tools by simply revising the approach to the sliding window.
 
Maxim Dmitrievsky #:
The properties of fractals allow you to get longer correlations, for the duration of the patterns. But it is necessary to adjust the windows. In essence, the method is to isolate such states by dividing the series into 2 parts and subtracting one from the other in reverse order, reversing chronologically the second or first part, i.e. mirroring the graph at the point (attractor). These phenomena are what manifest themselves in Fat Tails and memory. That is, the pieces of the graph on the right and left of the attractor are strongly correlated.

The process is formed simply: 1. Bifurcation point, past pattern breaks, new influencing information enters the market. 2. After the information is partially already accounted for by the market, an attractor is seen to predict the remainder. 3. Collapse again, the previous dependencies stop working. The beginning of a new bifurcation is also partially predicted.

There is no such thing in classical econometrics. But it is possible to use its tools, simply by revising the approach to the sliding window.

I don't get it. Is it possible to identify strong bounces by subtracting the right part from the left part relative to the centre and divide them into groups? Why longer correlations? These are short processes. How about more obvious, stronger ones?

In general, it may work with time or event binding. But the window size should be trained to get it somehow.

And the formalisation of events has not been solved. Only time, apparently.

 
Maxim Dmitrievsky #:
The properties of fractals allow you to get longer correlations, for the duration of the patterns. But it is necessary to adjust the windows. In essence, the method is to isolate such states by dividing the series into 2 parts and subtracting one from the other in reverse order, reversing chronologically the second or first part, i.e. mirroring the graph at the point (attractor). These phenomena are what manifest themselves in Fat Tails and memory. That is, the pieces of the graph on the right and left of the attractor are strongly correlated.

The process is formed simply: 1. Bifurcation point, past pattern breaks, new influencing information enters the market. 2. After the information is partially already accounted for by the market, an attractor is seen to predict the remainder. 3. Collapse again, the previous dependencies stop working. The beginning of a new bifurcation is also partially predicted.

There is no such thing in classical econometrics. But it is possible to use its tools, simply by revising the approach to the sliding window.

I did not consider fractals, but looked at the chart.

 
СанСаныч Фоменко #:

Same RSI, but with some nuances.

Once again: it is not the list of predictors that is important, it is the algorithm of selecting these predictors that is important. When the window moves, the mentioned RSI may or may not be used. This is the problem, and the first of a whole set of problems.

Methods of averaging parameters of a series are generally understandable, the logic of their creators is also clear (not always and not completely of course))). ), indicators are created on this basis. The reason for the lag is clear. And the possibility to remove the lag is not clear.

I do not know, there was an idea here (not mine))) ) to generate rules from price indicators and indicators and see the result by signals for trading.

But this is not a meaningful re-selection / selection of signs.

It ispossible to make signs from prices and their averages of neighbouring currencies.

In general, I do not understand the selection algorithm yet.

Clearly it is levels of extrema, tick speeds, trend, width of the trend corridor, frequency of price returns to the corridor boundaries, on different time scales.....