Statistics as a way of looking into the future! - page 17
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
There are. I have already given them to you several times and I will not repeat them. If you don't see them or don't want to see them, I can't help you.
Understand, no one here is proving anything to you and no one is obliged to do so. You asked me to explain it to you, I tried, but you didn't understand. Alas, not everything is so simple in this cruel world. Not everything can be done in one fell swoop, you have to have a certain knowledge base.
There are. I have already given them to you several times and I will not repeat them. If you don't see them or don't want to see them, I can't help you.
Understand, no one here is proving anything to you and no one is obliged to do so. You asked me to explain it to you, I tried, but you didn't understand. Alas, not everything is so simple in this cruel world. Not everything can be done in one fell swoop, you have to have a certain knowledge base.
And what, you managed to predict the price within the framework of Dynamical Systems Theory not by leaps and bounds and with a certain knowledge base?
I think a good TS cannot be built without making a forecast, let it be 0.62 rather than 1, it means I enter the market with SL=TR in 62 trades out of 100 and get a guaranteed profit.
I cannot do it without a forecast, otherwise I might lose my head in a rage.
I think the author has expanded the concept of prognosis to the heights of space and then everything that a trader does in the market is his forecast, but then bad TS are based on the forecast too, or he has narrowed this concept down to indicator systems, but then good TS are also possible without a forecast, I thought and thought what to call non-indicator systems without an initial forecast and came up with SIMETRIC))), but maybe they already have a name?
And what, you managed to predict the price within Dynamical Systems Theory without jumping in and with a certain knowledge base?
If I answer yes, you will demand to show you proof until I renounce posting on this forum altogether :) So I will answer "no".
If I answer yes, you will demand to show you proof until I renounce posting on this forum altogether :) So I will answer "no".
Your rationalisation about "what to answer" is just a reinforcement to help you admit the truth - neither you, nor anyone else on this forum, nor Anischenko, nor the founding fathers have positive results on price prediction. Such is the cruel world - those who can't think, they gobble up everything like a chicken in the hope of finding the seed. And one only needed to read about the limitations of the Dynamic Systems Theory application, so as not to winnow.
If so, I'll be the first to throw away all my experience with NS and join Prival as an apprentice! Right now (or almost right now) I'll start to reread the thread about the Flow in Forex and build a Kalman filter.
The only pity is that I probably won't have to do it. And the reasons, I hope, will become clear soon.
You should not compare linear regression and neural networks, each method has its advantages and disadvantages. For example neural networks give a smoother signal model and better phase response, less lag of 1 - 2 bars when forecasting but linear regression gives more stable signal far beyond the training. The figures show an example of modelling from the same raw data using neural networks and linear regression. The range of quotes for model training is taken from May 20 to June 10, the range of rate fluctuations in this interval was from 1.54 to 1.6 . Yellow and pink signal are neural networks trained on the same input data but for different target functions, red and blue are linear regression trained on the same data and for the same target functions as the neural networks, i.e. yellow and red for one target function and pink and blue respectively for the other. Figure 1 shows the graphs within the range where the training took place. Fig. 2 shows graphs outside the learning range, as seen in Fig. 2, from August 8, models on neural networks began to produce a large error, i.e. training was enough only for 2 months because the rate was lower than 1.52, while the lower boundary of the training sample was 1.54. Fig. 3 displays charts with quotes until October 13, it's clear that models based on neuronets show strong distortions, meanwhile models based on linear regression preserved their stability without re-training during the very volatile market. I combine both neural networks and linear regression, weakening the weaknesses of each method and reinforcing their strengths.
There is no need to oppose linear regression and neural networks, each method has its advantages and disadvantages.
Am I arguing with you?
Of course a comparison of methods has to be made in the context of the task at hand. For me, for example, one-step-ahead prediction with prediction at each step is relevant. In such a formulation, NS is probably out of competition.
The data you presented is curious. Unfortunately, the quality of the pictures is not excellent, and it is difficult or even impossible to see anything on them. If possible, zoom in the first third of the second fig. - I want to see the quality of muvings in the area close to the boundary of the last optimization of NS. You can also present the data in a more informative form - in the form of predictive cloud in coordinates of price increments and muvinings (see page 3 of this thread).
Unfortunately, the quality of the pictures is not great and it is difficult or even impossible to see anything in them.
The pictures can be clicked on and are then shown at their original scale.
I have a question for Pilligrim: what is the input vector for these models and what is the output? Without this data, these figures don't tell you anything.
Pictures can be clicked on, then they are shown at their original scale.
No, it's not impressive. Let him redraw it.