From theory to practice - page 580
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
OK. Let's put it down. I don't mind - I just want to line my own pockets and I don't care about other people's.
1. I work with ticks in a sliding second time window.
2. for example, take a window = 14400 seconds, and create 3 (three) FIFO(14400) buffers.
3. With frequency = 1 sec. count the difference between current and previous price value (increment). Everything in a row, no matter whether it was a real tick or not, is written into buffer #1. We calculate the sum of all values in it. It is the price. Black line.
4. Count increment modules - we write them into buffer #2. Count the sum. Divide by 14400. This is the average rate of change in price. Let's call it C.
5. Now it is a little bit more difficult. We need to count the number of real ticks in this window. At every step, we look whether the increment itself or the time of value arrival has changed. If it has, we write a unit (1) into the buffer №3, if not - 0. Count the sum of the units. For example, we get 12345. This is the real number of incoming ticks in 14400 seconds. The sum of increment units from buffer #2 is divided by 12345. This is the average value of Lambda increments.
6. Calculate diffusion coefficient by formula: D^2=C*Lambda*T. Standard deviation Sigma=sqrt(C*Lambda*T).
7. Now make the assumption that all increments in BP are weakly dependent. The sum of such values gives a number belonging to a normal distribution.
6. From zero we plot support/resistance lines = +-2.5758*Sigma, where 2.5758 is the 99th quantile of the normal distribution. These are red and blue lines.
7. For the price it is the same, only +-2.5758*Sigma is not taken from 0, but from the initial reference point, i.e. the first element in the FIFO(14400) buffer.
That's it. This is the maximum that can be squeezed out of standard (not anomalous!) diffusion.
Alexander! If I unload three columns (Sum of increments and variance channel) can you substitute it to see the graph? Because I work with online exel with a limit of 3000 cells.
I'm a bit lazy, Gianni... Diffusion hasn't brought me a penny personally. Lying in a roadside ditch with empty pockets, I'm writing from here... I'm waiting for the Person in this thread to show the Result and breathe hope.
They burned the grail!
A-ya-ya-.... Where? Where to go for happiness? Show me the state based on this algorithm, please... I've had +5% profit in six months. Something's missing.
A-ya-ya-.... Where? Where to go for happiness? Show me the state based on this algorithm, please... I've made +5% profit in six months. Something is missing...
Maybe you just don't have enough time!
Your thoughts and calculations seem quite adequate.
There is only one question that haunts me: why is the 99% quantile?
Are all the other levels exhausted or what?
I'm a bit lazy, Gianni... Diffusion hasn't brought me a penny personally. Lying in a roadside ditch with empty pockets, I'm writing from here... I'm waiting for the Person in this thread to show the Result and breathe hope.
In general, I get a multiplier in the variance = 5. And this figure is sort of reasonable, so I want to look at the long interval.
Perhaps you just don't have enough time!
Your thoughts and calculations seem quite adequate.
There is only one question: why the 99% quantile?
Are all the other levels exhausted or what?
:))) I don't know. It's possible to play around.
The state needs a grail, even with a small secret addition to this algorithm.
At least I will know that it is possible, get out of the gutter and continue the search.
I have so +5% profit in six months. Something is missing...
It's all about the quantile, of course... less than 999% is not even worth sticking your nose out. ))
Got a picture like this in 3,000 minutes
Alexander_K:
get out of the gutter and keep looking.
What about constructing such a theory, since we are looking for deviations from the mean and a return to zero.
Suppose we have two sums of increments.
A = sum of price increments
B = sum of average price increments
Calculate the variance C = A - B = the current deviation from the mean
Next, calculate the variance for A and B
Then sigma = Abs(dispersion A - dispersion B)
We build a channel and if C has broken through the level we assume the price will return to the mean.
So far I have only seen one order with euro on the 14th , I think it was at 16:01 buy Tp of about 70 pips.
Nah, I'm fed up with the theory... State wants the Man. I'll be waiting.