Machine learning in trading: theory, models, practice and algo-trading - page 2084
![MQL5 - Language of trade strategies built-in the MetaTrader 5 client terminal](https://c.mql5.com/i/registerlandings/logo-2.png)
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
If the frequency has the maximum amplitude, then it is easiest to isolate it from the signal and it will give the greatest profit, imagine the sum of the sines, one with an amplitude of 10, the other with an amplitude of 100.
imho, the ideal indicator is an oscillator (bandpass filter) tuned to the frequency with the maximum amplitude.
By the way, can you show how this filter looks like in work...
Maybe really those TCs are much cleaner to filter with harmonics...
this metric means nothing to bots
why measure with it )
I understand what you mean, and I agree, but let's forget about wizards and harmonics...
We need a universal way to produce the optimal parameters...
Imagine another TS, trading the MACD at the intersection of the zero line with the signal line. Will the optimal period of such a TS be synchronized with the harmonic frequency with the maximal amplitude?
In my opinion, no.
You can find the period on the spectrum, but you can not find a "bouquet" of several functions for TC
Macd is a bandpass filter + low-pass filter from it. By spectrum we get cut-off frequency - 2 parameters, we take a signal line arbitrarily, it adds smoothing and delay
The macd is a bandpass filter + low-pass filter from it. We get the cutoff frequency from the spectrum - 2 parameters, we take the signal line arbitrarily, it adds smoothing and delay
So, basically you can describe any indicator by a harmonic combination?
You don't need those indicators, you need the right harmonics to replace them, and if you build rules based on harmonics, you can model any system on indicators?
Why measure with it?)
just to look at
And can you show by the way how this filter looks like in work...
Maybe really that their those TCs, harmonics are much cleaner to filter...
why measure with it then )
Without looking into the future so-so
In fact, you can describe any indicator by a combination of harmonics?
You don't need those indicators, you need the right harmonics to replace them, and if you build rules based on harmonics, you can model any system on indicators?
The classic problem on this subject in our field is the Markowitz portfolio theory. There you get not one, but many optimal portfolios - the choice of a particular one is made on the basis of trader's preferences for the ratio of profit to its volatility.
The question is philosophical, high peak or plateau is much lower, or high density in a small point or average in a larger volume))) And when there are 5 parameters it's already complicated). Portfolio problem, on the one hand is multifactorial, on the other hand each portfolio outputs one parameter from time. This is still a different task from Mashka's period optimization for the best description (closest in frequency and amplitude) of the characteristics of the series.
I haven't got to the news, I can either get a price series and compare it with a news one or use the tester to parse the news one.
purely to look at
You take a sliding price window of 10-20
do PCA on 10 components
take each component and do PCA on it, but in a sliding window of + - 100
add it to the model and take your 0.7 - 0.75% acuracies
The classic problem on this subject in our field is the Markowitz portfolio theory. There we get not one, but many optimal portfolios - the choice of a particular one is made on the basis of the trader's preferences for the ratio of profit to its volatility.
Describing the price series. (for myself)
The trend cannot be eternal. A stable state of course. The probability of the term of stable states is higher in the range of +20% of minimum, -20% of maximum on sufficient history. At different scales, the series behaves the same way.
What is similar is temperature and price. We measure the temperature discretely in time, while it is continuous and there is a difference between measurements, and we know with a high probability that this difference is almost always less than some value X. The series is similar in that there is a difference between ticks. And we also know with high probability that this difference is less than some value X. Gepps and volcanic eruptions with tsunamis are also similar).
And these X's depend on the time scale).
This article has everything
Thanks, I read it... I don't understand everything, but that's my fault))