From theory to practice - page 1510
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
Comrades mathematicians wanted to ask you for a solution to a question:
visually you can see that the second picture(the red line) is closest to defining it as an exponential function.
I certainly understand that we can calculate the rate of change of the numerical series represented by the red curve.
but as we can see the velocity in both graphs grows, but only the second graph grows exponentially.
How can we mathematically calculate how close the curve (or the numerical series represented by the curve) is to an exponential curve?
How do I mathematically calculate how close the curve (or the numerical series of the curve presented) is to an exponential curve?
Make an array with the curve and an array of the same time with the reference curve (which is exponentially plotted) and calculate the correlation.
Wrong?
The pound, the devil, is tearing me up like wet toilet paper again...
What about diffusion and reading the whole book?
There was a lot of shouting about the grail.
Did the grail break before the owner's pockets were filled with money?)
What about diffusion and reading the whole book?
There was a lot of shouting about the grail.
The grail broke before the owner could fill his pockets with money?)
Alas... I'm going to church... The parishioners won't hurt me.
Make an array with a curve and an array of the same time with a reference curve (which is plotted against an exponent) and calculate the correlation.
Wrong?
That's an option by the way) I'll give it a try) Thanks!
The main idea was to try to formulate a dynamic cocoon model (e.g. at least in the form of y=kx+ax^p) trend component of which will depend on the current slope of distribution of the given area, then average all cocoons and try to find their common boundary, eventually such cocoon will not expand infinitely but will represent something like an amorphous snake/trough that sometimes swells, unfortunately the task is quite extensive as many questions appear, in the simplest case as you exactly pointed out this On a plot sometimes one looks better sometimes another, we can also use the law of the repeated logarithm, but there appear troubles in the form of a shift of the plot by two points and undefined values at zero, and it seems that nobody in the world is bothered by it, but visually it is not nice somehow, not neat, I suppose that the concrete form of the law is not even important, because in practice there will be some errors/slips, there is also such moment: we can simply increase the scale factor before the root so you might not make it to the factory in time...
In fact only daily rhythm is stable, other harmonics are visibly floating and, what's worse, they stratify and decay... I don't know what to do about it... There is only one thing for sure: looking at the news calendar one can suppose the "Joker effect by Peters" - a memory shift and a sharp fall in autocorrelation... Interesting options were suggested here on the forum too, lots of options, there's not enough time to try everything... From a purely visual point of view the movement should be no less than 200 bars of the working timeframe with a glance at the point of the last extremum and the scale of movements on the left in the history, but it is quite subjective, I understand how it sounds... So far I got used to orienting on the chart form itself and on the calendar...
As for earning purely by CPT from a random process (no memory, no Markov's, no all that) - it seems very questionable to me, I don't even understand how it can happen, maybe there is a special process that is not quite random, because the very concept of randomness, if you ask Shiryaev and Kolmogorov for a historical reference, is a process that cannot be said to be in any order = no pattern of dependence of specific outcomes = no algorithm that could be used to reproduce it, even partly I can suppose that it has something to do with a consequence of the Hinchin and Shiryaev DSTP and then we can specify a neighborhood S(n)*sqrt(ln(ln(n))))+e which will be a boundary which the process touches only a finite number of times? - But there are assumptions about stable variance and average, and unfortunately the market is not like that... And even logically it is not clear how it can work. We made such an experiment in one secret secret sect: we took generator data and built gradients after trend, for random IID data (and for resampled market data too) there was no shift in any direction on any scale only flatline, But for the market data we observed a soft continuation (arc slightly upwards) and then a rollback to the initial level or slightly below (arc goes down and gradually turns into an asymptote), that is, there was a visible effect of difference between the generated data and the market data by over 9000 averages, in a word, if it is possible to earn on the bare DTT, I am perplexed, magic indeed...
https://www.mql5.com/ru/forum/286022 look through it carefully and unbiasedly, discarding the shackles of ingrained stereotypes and prejudices ;) -- and understanding will come.
The pound is tearing me up like wet toilet paper again...
What's up with the sneakers?
the pound is the last resort, the ultimate in your strategy.
Alas... I think I'll go to church... Parishioners won't hurt.
So it turns out that if we traded the amount of increments everything would be fine, but we trade the price.
And the price increments in the sliding window do not correlate with the incremental sums (increments).
So it turns out that if we were trading the incremental sum, everything would be fine, but we are trading the price.
And the price increments in the sliding window do not correlate with the incremental sums (increments).
is the sum of the increments not the price itself?
We can say that it is the price, but we don't know the period. We can know the period if we have enough history, but not necessarily, because it is not a zero-start.