Using spline(s) - page 2

New comment
 
mladen:
dasio

Frankly I did not work on that. At first I thought to make it a multi time frame interpolation (which is already done in one version of "averages" where quadratic spline is used for interpolation) but from cubic on, the changed bars back would extend too much in my opinion.

Probably some other mode of finding points than equidistant (multi time frame is equidistant in it's nature) that would then be connected by splines should be used. since it largely depends on those points, it seems to be the crucial point that would determine the whole result (and usefulness) in the end

Thank you for your reply mladen.

So for you is unseful use highest high and lowest low for find point for the indicator?

 

...

daso

I don't use it that way (the minimum and maximum way)

Since this is a "manual" nonlinear trend line, I choose points that I find significant. If there is a trend, then I try to estimate where is the "continuation point" for a trend (which is in those cases not even close to extremes or peaks). If there is a significant change, I wait till I see some kind of confirmation that the change is going to last and then I adjust the points again. And then the extrapolation take over giving what the mathematical expectation (since it is simply following the mathematical model from the past values) would be. I might or might not agree with it, but sometimes the extrapolation is like a magic and sometimes it is misleading, so, like with any extrapolation, extreme precaution is advised

dasio:
Thank you for your reply mladen. So for you is unseful use highest high and lowest low for find point for the indicator?
 
mladen:
daso

I don't use it that way (the minimum and maximum way)

Since this is a "manual" nonlinear trend line, I choose points that I find significant. If there is a trend, then I try to estimate where is the "continuation point" for a trend (which is in those cases not even close to extremes or peaks). If there is a significant change, I wait till I see some kind of confirmation that the change is going to last and then I adjust the points again. And then the extrapolation take over giving what the mathematical expectation (since it is simply following the mathematical model from the past values) would be. I might or might not agree with it, but sometimes the extrapolation is like a magic and sometimes it is misleading, so, like with any extrapolation, extreme precaution is advised

Ok thank you. Kow i must know wich are the point that you mean. I don't understand. Can you explain me with definition or image?

Thank you

 

...

Here is one example : as you can see the last points are move and I even moved the last point of the line in the future (in this case I did it more for example sake then as a kind of my expectation at a moment, but I hope this can explain a bit more how I use it). This way, see what the extrapolation is showing what would be the "logical" continuation of what is on a chart :

So, it is a kind of tool that we could call a "plot an pray" Without kidding, it helps me visualize what are my thoughts about the market conditions and expectations. It is purely a discretional too that helps me in my analysis

dasio:
Ok thank you. Kow i must know wich are the point that you mean. I don't understand. Can you explain me with definition or image? Thank you
Files:
spline.gif  29 kb
 
mladen:
Here is one example : as you can see the last points are move and I even moved the last point of the line in the future (in this case I did it more for example sake then as a kind of my expectation at a moment, but I hope this can explain a bit more how I use it). This way, see what the extrapolation is showing what would be the "logical" continuation of what is on a chart :

So, it is a kind of tool that we could call a "plot an pray"

Without kidding, it helps me visualize what are my thoughts about the market conditions and expectations. It is purely a discretional too that helps me in my analysis

Ok so you have to try to follow the price in the market finding the point wher you think it will change. Doing so the indi extrapolate what the market would be do in the future. It's right?

 

...

Yes, but in the example above more or less that end point is the end of Friday, so I will have to rethink till monday what to expect, but as far as I see it is what we should expect (so even like an example it was not bad)

In estimation the only "problem" is that it is men estimationg the bigger part, and computer the smaller part. But then it is probably a matter of preferences : manual or automatic. This one has parts of both worlds and maybe that is what makes it different when compared to any other similar ones and maybe that is the exact edge that we should keep - the combination of our (human) estimation and computer taking over with extrapolation from that point on

dasio:
Ok so you have to try to follow the price in the market finding the point wher you think it will change. Doing so the indi extrapolate what the market would be do in the future. It's right?
 

Hi mladen, I wonder whether the spline indi could be combined with the Hodrick Prescott indi, e.g. to automate the spline indi by using past HP values and using the more actual values of the spline indi to overcome the HP end point distorsion problem.

 
Boxter:
Hi mladen, I wonder whether the spline indi could be combined with the Hodrick Prescott indi, e.g. to automate the spline indi by using past HP values and using the more actual values of the spline indi to overcome the HP end point distorsion problem.

I am not sure what would the results be

In both cases we would have extrapolation for that last segment (depending where would we place the "control points" on HP in order to define a spline) so both would necessarily be approximations. Have to do some tests to see what would the results look like

 

Hi mladen, here is an interesting implementation of my idea but using a digital filter first before fitting a cubic spline for forecasting:

AFIRMA = Autoregressive Finite Impulse Response MA

AFIRMA - MQL4 Code Base

This indi could be also very interesting for a mod of the Detrended Price Oscillator DPO.

 

Some Google Chrome translated english explanations of the AFIRMA indi:

"...AFIRMA (Autoregressive Finite Impulse Response MA) based on the digital filter accurately shows the price movement, though with a time lag. Moving Average based on the smooth function fitting (a polynom or sin/cos like in Fourier transform) has no lag but often shows the price movement not so well.

The proposed moving average smooths out the price movement using the digital filter for all candlesticks, except the latest ones, the number of which is equal to the filter time lag. These candlesticks are smoothed by cubic spline fitting using least square method. The condition of combined moving average and its first derivative continuity is set at two moving averages crossing. As a result, we have the smooth moving average that accurately tracks the prices without a time lag.

The input parameters are:

- Periods - mounted bandpass pass filter, which equals 1 / (2 * Periods);

- Taps - the number of delay/lag candles in the filter (must be an odd number), i.e. the filter must have Taps prices to form the new smoothed МА value;

- Window - selects an approximation of the filter:

Window = 1 - rectangular window;

Window = 2 - Hanning;

Window = 3 - Hamming;

Window = 4 - Blackman;

Window = 5 - Blackman-Harris.

On the chart, the first part of MA, built a digital filter, represented by a blue curve. The second part of the MA, constructed by fitting a cubic spline curve is depicted in red.

123
New comment
Reason: