Market phenomena - page 70

 

The raw data is the same, only now the histograms are in 0.00005 and 0.0000001 steps The phenomenon is many times stronger! :-)

P.S. The ordinate axis is percentage (sum of all histogram rectangles - 100%).

 

The fine structure of the phenomenon (all the same, only the abc axis shows an interval of -2 to +2 pips, I use this concept, although it is also about the logarithm of the price).

Such a fine structure claims to be a separate, second, phenomenon. What is the nature of these "forbidden zones"???

 
Svinozavr:

- Believe me, you're overreaching.
- That's typical of me.

// Pokrovskie Vorota

Kindergarten. To be honest, it is!




Found a phenomenon! one of the forex market phenomena is Svinozavr! ))
 
Dr.M.:

The fine structure of the phenomenon (all the same, only the abc axis shows an interval of -2 to +2 pips, I use this concept, although it is also about the logarithm of the price).

Such a fine structure claims to be a separate, second, phenomenon. What is the nature of these "forbidden zones"???

It looks like a fountain dance )
 

Third phenomenon: note that the height of the maximum does not change with step change (about 1.1%). At steps of 0.01, 0.001, or, here, in the picture below, 0.0005 from w=0.0001.

While the heights of all the others change(I build a histogram normalized to 100%, that is, the heights of all the rectangles, which become more and more as the step decreases = 100%).

 
Dr.M. 05.02.2012 11:06

Found the phenomenon. I am pleased.

Let's take EURUSD5.prn with at least 100 thousand points. Let's take the logarithm of klose prices. And plot the distribution not for price increments but for price logarithm increments. We'll see a Gaussian. No surprise there. Everyone knows that the distribution of price increments is lognormal, and it is clear why the logarithm price increments are distributed normally. But take a look at the picture in the appendix. Let's build a histogram with the step 0.0001 (there is an argument in the fraction w=0.0001 of the operator Hist)) - Gauss. And let's build it with step 0.000001 - what's that huge maximum there in the centre?!?!!



Is that really Gauss in those pictures? I don't recognise him in make-up.
 

No desire to write code to approximate the Gaussian (or rather look for it, it was somewhere), so I just drew a Gaussian, with a mean of zero and sigma 3, here on the same graph it is superimposed in blue.

Well... almost a Gaussian :-) so. there is some difference in shape. a small one. The tails are heavier. But that's not surprising. What is surprising is the phenomenon in the centre that occurs as you look at the decreasing pitch of the histogram.

 
Dr.M.:

What is the nature of these "forbidden zones"???

maybe in the Galton board the nails aren't there? ;)

What is the price discreteness? And how does it relate to the breakdown into frequency bands?

Are you working with 5 digits?

 
Dr.M.: Everyone knows that the distribution of price increments is lognormal

Maybe you and almost everyone else knows it, but not us. We think it is not Gaussian, but rather exponential. At any rate, the tails are more or less consistent, they are heavy and fat.
 
Mathemat:
Maybe you and almost everyone else knows this, but not us. We don't think it's Gaussian, but rather exponential. Anyway, the tails are more or less consistent, they are heavy and fat.
you mean two almost symmetrical exponentials?