HOW TO TRADE FOR NOTHING ? - page 3

New comment
 
gip: That's right Richie, the technique is useful. Only as a matter of principle, take non-harmonic functions, even for practice. Take some piecewise linear ones. Harmonic ones can lead you down a blind alley.
Tangents or something? Seriously, which ones? I thought Fourier was more appropriate.
 
Swetten:
Is a Fourier expansion a periodic function?

Oh-oh! A Fourier series can be used to express (approximately) a linear, quadratic or other functions, on a given interval. The Fourier series itself is periodic.
 
EvgeTrofi: 1. Under what condition were the girls reversed? .......................
Eugene, I understand your sense of humour, but there is no time for girls now, it's hot.
Prival : this is a strange question... It's an absolutely periodic function. I build a spectrum, even without seeing the formula, I pull everything I need from the spectrum, the number of components, their amplitude, frequency and phase, I get an analytical function, ... at the kinks I flip...
We need something simpler and more reliable. Most importantly, reliable. Like this straight line on the graph, for example.
 
Richie:
Tangents or something? In all seriousness, which ones? I thought Fourier was more appropriate.

I'm sorry. It's Friday night. I'm not thinking straight in the heat of the moment. Didn't mean to intrude. It's just that you're starting to get snippy, and it's the right approach. It's only misguided. The main thing is not to assume that the market is the way you modeled it. Don't carry a one-size-fits-all, tried-and-true apparatus.
 

here is the solution

I put it in Fourier, graph 1. the output is a spectrum. this is figure 2. even without seeing the formula, I can determine clearly from the spetter,

1. There are four components. 4 red sticks on the graph

2. These sticks are located at frequencies 7 10 15 and 24

3. The amplitudes are 8 5 1 and 7.

Now you can compare it with the formula at the top

 
Prival:

Sergey, a small off-topic question. On a real graph, how can you significantly increase the accuracy of the calculations, using the Fourier method? By increasing the number of "cosines" in the equation? By feeding the Fourier price through a filter, would an average MA with a small period be suitable for this?

 

there is a such a story. fantastic. like a superior race left a robot in space that knew the answers to all questions. people stumbled upon it, assembled a whole expedition, flew and arrived. they prepared for a long time and asked one question humanity will survive? which was answered, what is life? and it flew away, and only from afar did they hear that to ask the right question you should know at least 2/3 of the answer. this is me so hot.... Friday...

so i'll clarify your question because i don't understand

1. the actual graph of what ?

2. it is a formula and the accuracy of its calculation is most often determined by the digit capacity of the numbers represented. how to increase let's say 2/3=..... I do not know

...

try to phrase the question in an easier way so that a dumb military man like me would understand

 
Prival: Try to formulate your question in a simpler way ........

Tried extrapolating the EURUSD price using Fourier. The financial result is weak but positive, there are two drawbacks:

1. Low prediction accuracy; 2. Big drawdowns almost in every trade (low quality of trades);

Question: What methods may improve accuracy of price prediction using the Fourier method?

 

got it.

let me try to explain with the example above.

1. even if you add noise to 1 graph, you can still get those 4 sticks out of the spectrum (figure 2) (under certain conditions) .

2. the Fourier itself is the matrix apparatus not of prediction but of analysis of what is there now.

3. After all, in this example, I use Fourier to pull out the formula for what is in the input and that's it ... and then I predict using this formula.... the problem is that in your example, the function (its values don't change over time) knowing all frequency amplitudes and phases - they don't change, I can calculate what value this function will have in the future... that is, the forecast is not by Fourier, but by a formula=process model...

4. In forex there is no such thing, all these components change over time + the number of these sine waves may change...

5. We can use a Fourier transform to determine what is going on now, and just hope the market behavior doesn't change, by imposing this constraint we can predict into the future, but the problem is that the market changes, it changes all the time, if you look at it from the perspective of spectrum processing, the spectrum is always drifting...

 
Prival:

Here's the solution.

A clear example, thanks for the topic and I remembered my cadet years - when I tried at all costs not to learn these formulas and quadripoles in "Theoretical Foundations of Electrical Engineering", but was forced to, for which I am grateful

As for the subject: I tried to get into the maze of mathematics - I started to look for prognoses using derivatives and extremums of functions - i got semi-working indicators - i started to analyze their work and i concluded that the lack of Cartesian coordinates, or rather starting/zero points and binding to discrete time - Bars and discrete price - Close were to blame

so in this example - it is simple not because the clear harmonics of the function are visible, but because there is a clear binding to the coordinates - what to "bind" to in Forex?

Visually, we are bound to Cartesian coordinates in the timeframe - here you have time, here you have price - here you have "dummy" - and everything is clear, but there are no axes where the price chart would pass through zero points - as an option - an open market order that defines these zero points :)))

You may try to take the Pivot Point of the previous day as the current zero axis - but this is not a self-deception

123456789
New comment