Statistics as a way of looking into the future! - page 8
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
I don't understand what I'm supposed to see here. It looks like a trivial AR(1), i.e. if yesterday was down, today will be down, if yesterday was down a little, today will be down a little. Accordingly, the forecast is late with the price reversal and late with the acceleration/deceleration of the price. That is, if there is a reversal on the zero bar now, the forecast will show it only on the next bar.
If you mean ATR(1), here it is, compare it. If some other, give me the link.
I do not say if it is good or bad, it is just an indicator, which contains the model of the process (most probably incorrect) and there is a forecast based on this model
I was not referring to any particular indicator, but to autoregressive process lag 1 - https://en.wikipedia.org/wiki/Autoregressive_moving_average_model.
It looks interesting, but is of no practical value due to the flaws mentioned above.
Sorry to interject myself into the dialogue.
Can m_a_sim or Prival share their algorithms to study them (post here or send to my email in my profile ).
I am interested in statistics myself, but I approach it so to say from the other side, soon I will lay out here my software (with detailed explanation) for this topic. But I would like to expand its capabilities and explore other approaches you are now discussing here. Very interesting.
Personally, it fascinates me to "see" what will happen. I want to write an indicator.
Thanks in advance.
it's minutes, the shorter the forecast horizon, the more accurate it is. The graph is not drawn from Close, but from the "true price", its estimate.
Here is the chart, the red line is the forecast and the white line is the "true price" estimate. It (indicator) does not redraw.
It's not that simple, needs multi-currency analysis at the very least. And for that you need matrix operations. I still can't make an analogue of transposition as in matcad.
Z.U. It's too early to go into battle yet.
For example, I want to use the results of your analysis for the H1. I really want to feel the algorithm for its practical use!
Sergey, can you post the forecast cloud built according to your algorithm, but for H1? I really want to feel the algorithm for its practical use!
I am a bit confused about the cloud. Now I will make a file and post it (date, time, Close, estimate Close, forecast for this bar) for H1 for two currencies.
Z.U..
I'm ready to exchange this indicator (source), the library of matrix algebra
I need operations of addition, subtraction, multiplication, division (it's simple), the circulation of the matrix (there is a here 'The theory of random flows and FOREX', has not yet tested), transposition (here I have cracked my head, although the operation is the easiest) and calculation of the trace matrix(tr).
I need this MathCad procedure all symbols are matrices.
Gentlemen, don't panic. Why all the euphoria? :)
After all, you can see with the naked eye that there is no honey here yet.
Here's a little analysis on the first price chart that Prival posted. I digitized it to do a regression analysis:
digitisation
The scale on the value axis is pixels. There is no difference for regression analysis. The digitisation error is +/- 2 pixels, which is sufficient for our purposes.
Now let's build a regression of the closing price differences and the two curves (estimated and forecast):
regression
The graph on the left is the regression of the closing price difference and the difference of the estimated curve. The slope is 0.5720
The chart on the right is regression of closing price differences and forecast curve differences. The slope is 0.3183.
So, if we follow Neutron's methodology, a slope of 0.3183 on the volatility of a minute chart will give about one point, which with the spread taken into account means a true loss. Plus a big spread from the averages.
All in all, peace of mind. Let's wipe off the drool and get back to basics :)
Gentlemen, don't panic. Why all the euphoria? :)
So, if you follow Neutron's methodology, a slope of 0.3183 on the volatility of the minute chart will give about one point, which, taking into account the spread, means a sure loss. Plus a big spread from the averages.
All in all, peace of mind. Let's wipe our drool and get back to the basics :)
That's what I'm saying!
Are you sure(m_a_sim & Prival), that you do everything right, when you get a tangent equal to 1?
Once again. Such a value may be the result of an error in the implementation of the prediction algorithm. For example, you take the current price increment and correspond it with the current increment of your magic indicator (for example, a usual mask). You can be sure you will get the slope close to 45 degrees with a narrow smoothing window! This is a mistake. You should take the prediction of the indicator's increment and not the current increment itself!
Well, I don't know how else to explain it!!! Think a little bit about what you're posting.
I was not referring to any particular indicator, but to autoregressive process lag 1 - https://en.wikipedia.org/wiki/Autoregressive_moving_average_model.
It looks interesting, but is of no practical value due to the above mentioned flaws.
If I understand it correctly, these are regression models with a random component (which I believe is missing in m_a_sim model), here it is in Russian http://www.statsoft.ru/home/textbook/modules/sttimser.html#aarima
I did them and I put the ACF calculation in the (codebase) 'Autocorrelation function' as it (ACF) is one of the foundations of these models.
I am trying to model the pricing process differently. I use a system of stochastic differential equations. In addition to model errors, we can put measurement errors into it. What brokerage companies give us is not the "true price" but its estimation, at best we can hope that it lies in the middle of the spread with some probability. All of course IHMO, but this is how I look at this curve.
Well, I do not know how else to explain it! Think a little bit about what you're putting out.
Or just look closely at the chart. The "forecast" line is actually a copy of the price line with a shift of one bar. In other words, the forecast does not predict anything, but rather shows with a lag what has already happened.