Demonstrating the cluster approach to market... - page 17
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
I told you the truth: I don't need mash-ups and other smoothing tools in this system at all. But my calculations are quite different.
----------------------------------------------------------------------
How on earth do you refer to a chart for price without averaging ?
However, silence.... silence...
If I was an expert on tics, I would have done what I want a long time ago. It's not working. I want to build a multicurrency tick.
I also think it's better to use different mathematics when it comes to filters and mashups. I don't know what to believe though, I'm not a prophet. The only thing is if I were making filters I would use PF. It is by analyzing the spectrum that you can adapt the filter. Or rather the spectrum itself tells you what kind of filter you need. After all we've heard here in this thread saying that the spectrum is "floating", yes it is and it can't be otherwise. If it were not floating it would have long ceased to exist.
If it did, where would it go?) ?
The algorithm is as follows.
Let's take sample, preferably with maximum sampling rate (ticks). Let's take into account that sampling rate is not a constant. Smooth the frequency. Eliminate quantization noise. Build a spectrum, preferably DFT rather than FFT. Apply window, hemming, hemming ... etc. further investigation. Let's say we select 3 or 5 or 10 highest amplitude components of the spectrum and zero out the rest. Reverse Fourier transform, and we get a curve, plot it in the future, analyze its predictive properties (goodness of fit rule branch figure where 45 degrees) and in a circle until you find an acceptable result or, better, many results and a switch between (if the spectrum has ..... then work with this adaptive filter, if something else, then the next one).
Yes, fundamentally different. More likely even physics, if we are talking about cluster analysis. Well, I've got one more passage.
Semenych's indices are one of the very few in technical analysis which can be justified logically. Of course, there is an element of faith here too, but it is still not as great as with wizards, digital filters or Fibs.
Filtration of the price, which is something very close to a martingale, leads to a martingale. We only think or would like to think that a smoothed price shows us more of a trend or less of a flat. It is nothing like that. Still the same hopelessness.
And we won't do anything with this nasty martingale until we go beyond the analysis of one pair. This is where the individual near-martingale price processes begin to show us something of the Looking Glass, i.e. what is hidden when analyzing just one symbol. It is roughly the same as going out of two dimensions into three.
It turns out that analysis of the whole market, not just one pair, allows us to arrive at a logically consistent and indistinguishable "flat trend" classification. Moreover, it appeared to me recently that there are not even two qualitatively different types of flat trends, as I said before, but at least three. And in two of them it is possible to trade using flat systems, and in the third one - not possible because it is very dangerous. But trends are much easier than flat systems.
This classification follows from another mathematics than the trivial sum of codes which is mainly discussed in this thread. A simple sum of codes, even weighted ones, will not give us the answer to the question about how to draw the right border between a flat and a trend.
A simple sum of codes, even weighted ones, will not tell us how to draw the right line between flat and trend.
One can feel the approach to the understanding of the innermost.
For my part, in continuation of what I have said, I ask you to pay attention that the indicator given by me tends to zero on a given symbol,
(which means striving for the equality of "the cluster value of base currency" and "the cluster value of quote currency")
just on the time interval, when the instrument is in flat.
In fact, this draws the line between a flat and a trend on an instrument.
Flat on the instrument - the market (cluster) makes alignment of the cluster (full-market) values of the base currency and the quote currency,
thereby bringing the instrument out of the unnatural tension of the instrument with the full-market prices inequality
base/quoted currencies.
Trending on the instrument - for reasons unknown to us at present, an imbalance occurs between the full-market prices of the base/quote currencies,
which we can see on the chart of this instrument as a trend.
And here is the most interesting point - we try to play with the chart of this instrument in the most bizarre ways, we filter it,
build all sorts of levels, look for support/resistance, in short, let's try our best to predict the future fate of the nascent
of the emerging movement. And this chart of the instrument is just a form, in which the market has reflected the imbalance formed for the full-market prices
.
base/quote currencies imbalance.
No one can predict the fate of this movement. What is left for us to do ? Assuming it is fair to say that any imbalance in
that any imbalance in the base currency/quote prices will be eventually corrected by the market, then that's a good thing too.
Opening whenever this imbalance occurs and closing whenever the market eliminates it is a market strategy.
Trying to predict movements by means of various kinds of graphical constructions is a hopeless endeavour.
However, all sorts of tricks in the form of filters, etc., are nevertheless extremely necessary. They are necessary not for trading, because, as we have noted,
the chart of an instrument is only a form in which the full-market imbalance of base/quote currency prices is manifested.
They are essential for the correct and accurate assessment of the impact from base/quoted currencies on the rest of the market currencies.
After all, if we sum up the values of all the currencies in the market (cluster), we get zero at any given time.
Filtering the price, which is something very close to a martingale, leads to a martingale. We only think or would like to think that the smoothed price shows us more of a trend or less of a flat. It is nothing like that. Same old hopelessness.
This is because you associate filtering with suppression of high frequencies only, i.e. low-pass filtering.
How to get rid of constant component and infra low frequencies to get oscillator and drive it into a cluster without filters?
Take a sample, preferably with a maximum sampling rate (ticks). Note that the sampling rate is not a constant. Smooth the frequency. Eliminate quantization noise. Build a spectrum, preferably DFT rather than FFT. Apply window, hemming, hemming ... etc. further investigation. Let's say we select 3 or 5 or 10 highest amplitude components of the spectrum and clear the rest. Reverse Fourier transform, and get a curve, plot it in future, analyze its predictive properties (goodness of fit rule branch figure where 45 degrees) and in circle until you find acceptable result or, better, many results and switch between (if the spectrum has ..... then work with this adaptive filter, if something else, then the next one).
How do you align the frequency? And why DFT and not FFT?
How to get rid of constant component and infra low frequencies without filters to get oscillator and drive it into cluster?
How. It's just a waving machine with an ultra-long period. I don't think it's that critical. Or are the frequencies floating around here too?
I don't use oscillators either. The main problem with most oscillators is that they are not rationed. Artificial rationing by logistic functions like a hypotangent leads to saturation, i.e. insensitivity of an oscillator to variations near its "threshold" values. The price has reached a boundary level, you open and it keeps changing in the same direction. The artificially normalised oscillator doesn't sense anything and still shows that you should open the position in the same direction in which it was opened.
Making an unsaturated normalised oscilla is a whole art. It probably requires that the density function of its value distribution be as close to Gaussian as possible.
If I was an expert on tics, I would have done what I want a long time ago. It's not working. I want to build a multicurrency tick.
I also think it's better to use different mathematics when it comes to filters and mashups. I don't know what to believe though, I'm not a prophet. The only thing is if I were making filters I would use PF. It is by analyzing the spectrum that you can adapt the filter. Or rather the spectrum itself tells you what kind of filter you need. After all we've heard here in this thread saying that the spectrum is "floating", yes it is and it can't be otherwise. If it were not floating it would have long ceased to exist.
If it did, where would it go?) ?
The algorithm is as follows.
Let's take sample, preferably with maximum sampling rate (ticks). Let's take into account that sampling rate is not a constant. Smooth the frequency. Eliminate quantization noise. Build a spectrum, preferably DFT rather than FFT. Apply window, hemming, hemming ... etc. further investigation. Let's say we select 3 or 5 or 10 highest amplitude components of the spectrum and clear the rest. Reverse Fourier transform, and we get a curve, plot it in the future, analyze its predictive properties (goodness-of-fit rule branch figure where it is 45 degrees) and in circle until you find acceptable result, and better yet many results and switch between (if the spectrum has ..... then work this adaptive filter, if something else then next).
Yes!!! The details are all wrong, but correct in principle!
By the end of the year, I'll probably have finished such a multi-currency unit.
Yes!!! It's all wrong in the details, but right in principle!
By the end of the year, I'll probably have finished a multicurrency like this.
More details about all the wrong things. Preferably not just labels, but with justification.
Actually, it's a task for mathematicians to develop a "filtering" indicator which draws a "good" price line.
It's not even funny.