From theory to practice - page 887
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
This is clear without any research whatsoever. And for a long time.)
:))) And the choice of a moving average gives tangible improvements/deteriorations in channel performance. I managed to check that too.
So, in order not to randomly choose the coveted average - and we need a series of studies of sufferers. Let them work like the pope, and here they demonstrate the results. This is the only way to get close to the Grail.
I cannot make histograms, but I can only give you a file with price, standard average and "normally distributed average". I am waiting for the results.
The window is 1440 minutes.
No, Zhenya. You don't need a normally distributed average, you need normally distributed linear deviations of price from the average.
This will avoid the heavy tails that literally tear apart all channel strategies. This is a very big step forward + the weapon of kurtosis and asymmetry can reduce the number of losing trades. I've already checked it out.
I understand that. The file has a price and an average from which to calculate the deviations, and it is interesting to know if there is a normal distribution.
:))) I personally am not interested. I already know which MA is better than SMA. Let the eager ones like Bass and FelixWhite do their job - no need to sit in the corners like cockroaches.
:))) And the choice of a moving average does produce tangible improvements/deteriorations in channel results. I managed to check this too.
So, in order not to randomly choose the coveted average - and we need a series of studies from sufferers. Let them work like the pope, and here they demonstrate the results. This is the only way to get close to the Grail.
WMA is probably not the worst choice - equivalent to FIR filter. But coefficient selection is a real challenge. It's not so easy there).
You can hardly think of a better half-regression line, but the analysis interval must be relatively small.
And without rearrangement it is unlikely to work at all, either with FIR or with regression.
Well, there's a lot of room for creativity in the medium-sized structures.
Uh-huh. That's what I'm talking about. So, in order not to randomly choose this average, we need research.
In general, I must say one thing - Yuri Asaulenko, despite his nerdiness, is the best on the forum in terms of physical and mathematical understanding of the market, although, perhaps, he himself does not understand it :)
Uh-huh. That's what I'm talking about. So we don't have to randomly choose this average.
I understand correctly that the arithmetic mean, mode, median of the normal distribution are equal and the skewness with kurtosis = 0?
And we need to be inside the normal distribution.
Am I correct in assuming that the arithmetic mean, mode, median of the normal distribution are equal and the skewness with kurtosis = 0?
And we need to be inside a normal distribution.
Yes, that's correct. We need to be - but it still doesn't work. We are talking about asymptotic approximation and research should be done on archive data for at least 1 year. You should get a near-normal distribution of linear deviations from that coveted mean. Strictly normal will never be obtained.
So, the final average should show better asymptotic results than any other average.
So then there's nothing stopping us from being in this distribution with some margin of error without reinventing the wheel.
:)) How's that?
The problem has a very non-obvious solution and the aim is to minimise this error.
It looks good now - but at high emissions?
I cannot get out of my head Asaulenko's hypothesis that the truest moving average is the one whose linear price deviations form a normal distribution.
I set the sufferers a task:
1. to collect data on CLOSE M1 of any currency pair for a year.
2. to choose a rolling time window = 24 hours (1440 values).
3. Calculate linear deviations of price from moving averages:
a) median
b) arithmetic mean
c) a weighted average with different weights (time, absolute value of increment, probability of increment, etc.)
d) any other average of your choice
4. draw histograms
5. Calculate the central moments of obtained distributions
6. Demonstrate the results
Thank you.
But what property should we check on each pass?
What property tells us that the waveform goes in the centre of the channel?
is the distribution to check every time?
I wish I could build a program that would go through the periods of the MA and determine which one goes through the centre of the channel most successfully.
But what property should we check at each pass?
What property tells us that the macha goes in the centre of the channel?
what is the property that tells us that the mash goes in the centre of the canal?
Good thinking, mate. Only - not the MA period, the period we choose 1 time for ourselves, but exactly the type of average, which always goes to the centre of the channel. Yes, we should check the moments of distribution - asymmetry and kurtosis, which on average over many measurements have a minimum deviation from 0.
I personally spent a lot of time to find such a MA and don't want to tell you everything straight away. Although the SMA I had not the worst results, but it's definitely not the best option.