Predicting the future with Fourier transforms - page 54
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
Still fascinating, a glimpse into the future, albeit misguided :)
Cool ava))))) our man.
https://www.mql5.com/ru/forum/118912/page5
I wanted to clarify this. I tried to make it look like a blue line. To do this, I first need to extend/stretch it vertically (green line) and then move it backwards (blue line).
But the blue line will become shorter at the right edge, but you may try to continue it in the future through H and K. By predicting not the inflection points, but their displacement due to inertia.
This is not a price forecast . it is a forecast by H and K, which should reflect the inertia properties after certain price decompositions by frequencies.
I suggest that all of Valera's posts be moved to a separate thread called "Freudian Fourier Analysis".
He has polluted a good topic.
Freud:
Great, it's about time you sold your artistic masterpieces. No offence :)
)))) I understand, but I can't help it, with my crabby ability to express myself in a scientific style. I'd spend half a day explaining, but it's so obvious.
A word or two off-topic. I came across a thread on the internet.
Zigzag universal with Pesavento patterns http://www.onix-trade.net/forum/index.php?showtopic=373&st=75&p=204770&#entry204770
Sometimes the beginning of a pattern is smeared over time. spectral analysis can be used as an attempt to gather this smeared useful divorce into a pile, thus reducing the time lost waiting for the smeared beginning to be complete, due to the properties of the spectrum itself.
Cool ava))))) our man.
https://www.mql5.com/ru/forum/118912/page5
I wanted to clarify this. I tried to make a blue line look like a blue line. to do this, you first have to stretch the mask vertically (green line), then shift it backwards (blue line).
But the blue line will become shorter at the right edge, but you may try to continue it in the future through H and K. By predicting not the inflection points, but their displacement due to inertia.
it's not a price prediction . it's a prediction by H and by K, which should reflect the inertia properties after certain price decompositions by frequencies.
Please excuse me if I have misunderstood or missed anything.
Some of my private thoughts on Fourier analysis and why it was not very successful:
1) Trading robots are the main contributors to the high frequency component of the signal spectrum. It is meaningless for a trader to compete with them in this range, so the tick analysis is not attractive, especially taking into account that it does not contain information about buying or selling and who made it. Therefore the analysis should start at 3-5 minute intervals. But trading robots have an absolutely wonderful property of determinism from a theorist's point of view, as long as the trading robot works its dynamic characteristics are unchangeable. If there were only trading robots in Forex, it would be a researcher's paradise.
2. I consider High and Low signals to be characteristics of the high-frequency bar part and therefore omit them altogether. Thus, the variables Time, Open, Close and Volume are significant for analysis. It is from their ratios that the designer should derive the functions being analyzed. I did not see them.
3. Unfortunately, not only trading robots work on Forex, but people from the East and from the West. That is the biggest obstacle for prediction. But it is not all bad. For example, the start of trading on the Russian exchanges is a well-predictable event in the same way as the start of trading on the US exchanges. Both are from daily forecasts, so the minimum interval for most well-predicted events is a day.
4. If it does not make much sense to analyze the high-frequency component, then the first significant harmonic in the forecast should be at least twice lower than the investigated one. That is, if we want to confidently predict a 10-minute interval, then the interval under study should be no worse than a 5-minute interval.
5. Any sinusoid has that "bad" property, that it cannot be unambiguously plotted by three points. That is, if we investigate 5-minute harmonics, two notches of the beginning and the end of the sine wave will tell exactly nothing about its amplitude. Therefore, in order to confidently examine a 5-minute sinusoid, it must be scanned at least a minute interval.
6. Similarly, it is quite ambiguous to plot a 10 minute sine wave from two 5 minute points. So my 3 minutes and a day is not taken from the ceiling, it is in my private opinion a prerequisite for correct logging. That is, decomposition of a day into 480 intervals seems the most acceptable to me. Moreover, I think that trader's behavior on Monday is different from that on Friday. That is, I would add the daily reports by days of the week.
7. Any Fourier analysis will give the spectral content of the function in question. But the sinusoid has another "bad" property - the beginning and the end of the sinusoid are at the same level. Hence it is necessary to recover the constant component of the signal. Personally I approached this question again in terms of Fourier decomposition, which more than daily harmonics is quite correct represented in the daily interval by a sloping line passing through two known or calculable beginning and end points. This, in turn, allows studying long-period harmonics completely separately and in parallel with intraday harmonics. We give a daily timeframe and investigate a year from ... and up to the given day, of which we are interested in the prolongation of just the next day. Obviously, we also regenerate the constant for the year by dividing it by 254 for every day. And for each of the 3-minute intervals all this should be divided again by 480.
It seems to me that this should give a tangible result.