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
Here it is, our everything)))) has long been interested in such pictures, but I haven't built one myself, and the colour gradation is super))), if it's no secret, what from what and how?
Here it is, our everything)))) has long been interested in such pictures, but I have not built them myself, and the colour gradation is super))), if it is not a secret, what from what and how?
On the very first picture I have set the wavelets. on X samples, on Y on the left - frequency, on the right - period
... if it's not a secret what from what and how?
We can.
We will look for the pairwise correlation coefficient between neighbouring samples of the time series. For the selected time frame we have one coefficient in the range from -1 to +1. The coefficient value less than zero indicates the presence of antipersistence between samples, greater than zero - persistence in this TF, close to zero - get out of here! In its turn, persistence serves as an indicator of trendiness/collapse of the symbol on the selected TF. The last property of BP allows to use adequate indicators of the TA.
The correlation coefficient is in a window of n - samples. In this case we used Minutes for 2010 and by thinning them we have built the artificial TF from 1 min to 100 min. n was taken as maximal (how many samples in a year). For each TF we found correlation coefficient and plotted the dependence of this value on TF. I meant exactly this dependence in the quote above.
Fig. shows the found dependence of the pair correlation coefficient for different instruments at different TFs. You can see that almost everywhere the coefficient is negative indicating that the price tends to return to its initial value after the disturbance. This property is more or less characteristic of all symbols and is most clearly seen at small TF (see fig.). I used Alpari's data of 2010.
The question is what counts as "close to zero". For estimation, we can multiply the correlation coefficient at the selected TF by the instrument volatility in points at this TF and compare the obtained value with the brokerage company commission (also in points). If it turns out to be larger than the spread, then you will not succeed anyway, because the market is not an ergodic system and as soon as you open a position, everything will change for the worse (only for you).
What is wrong in your reasoning?
Correlation is a very sneaky thing. If there is a deterministic component in your quotes, you should be very careful with the results of correlation, because the deterministic component "plugs" the noise component, and we cannot judge about the statistical characteristics of the quote.
Let me give you an example I mentioned many times on other occasions.
Let's single out the deterministic component using the HP filter
Below, the "cyclic" component is an unfortunate name I think, I prefer "a million Pinocchios in a field of wonders", but the serum is there.
Let's take a closer look.
We see that this "cyclical" component oscillates quite regularly around a smoothed curve, which has an analytical, i.e. deterministic form.
Let us calculate how often the top and bottom of this HP curve
Negative more often than positive. But the trend has been falling and may be the result of this
If we start the analysis with a falling trend, we get a slight, but increasing negative values
If we take the rising British pound:
we get a different picture:
Which confirms my point that a higher number of negative or positive deviations indicates a trend in the plot.
Apart from the ACF tending towards the delta function, how else can noise be identified?
With ACF and delta functions you can prove anything
That's my point exactly.
I can't figure out the phase.
Shouldn't it change from -90 to 90 degrees? Why only to -54?