Searching for market patterns - page 105
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
Предлагаю поиск закономерностей также вести несколько в другом формате, а именно попытаться создать, придумать, вывести, угадать, ..... такую функцию, которая могла-бы удовлетворительно описывать заранее известную закономерность, пока заданной известными функциями или их сочетанием. Если предполагаемая функция рынка (услоно ее так назовем) справится с поставленной задачей, только тогда проверить ее на фактических рыночных данных. Остановимся пока на монотонно изменяющихся функциях без изломов. Любой может мне задать череду из 20 цифр, связанных между собой, известной только автору, закономерностью по предлагаемой методике. Попробую 10 из них использовать для определения параметров модели, а следующие 10 - для форвард теста. Функция рынка, если таковой вообще удастся получить, должна справиться со всеми функциями в любом их сочетании, поскольку рынок еще сложнее.
Congratulations! You have discovered the Neural Net - the formula is the same and works with any type of data! :)
"Let's stay with monotonically varying functions without kinks for now".
Explain the reasons for this choice, maybe it's just force of habit (working with smooth functions)?
It seems that without using switches just "ironing" and successfully predicting jumping quotes is hardly a dream come true...
IMHO......
The absence of a "market function" does not at all mean "exceptional market perfection".
The task is to attempt to create a "Market Function", which we shall conventionally denote by R(t) or simply R. This function will have to satisfactorily describe any regularities on the basis of known functions and their combinations. Anyone can propose his or her own version of the function R, which will be comprehensively tested for strength and reliability. Let us imagine, for instance, that I have my own version of this function whose form I don't want to divulge yet to avoid premature attacks. I would like to disclose what I know at this stage: It adequately describes the exponent, power and exponential functions, the branches of hyperbola and parabola, easily transforms to a straight line, describes any combination of the above functions combined into one function by their addition (subtraction) or multiplication (division), including the sine wave. For example, if the raw data is given by a pattern y=a+sint, then R "turns" into this function. We will continue the discussion on this topic if participants show interest in this direction of research, or point out another way to find market patterns.
... But there may be some perceptual inconsistencies here ;)
For example, in continuation and development of the idea presented by me here https://www.mql5.com/ru/forum/133625 (and, I note, practically not perceived by anyone), I make the next step in the direction of using predictive models - Model Predictive Control (MPC)
This approach began to develop in the early 60's to control processes and equipment in petrochemical and energy production, for which the use of traditional synthesis methods was extremely difficult due to the extreme complexity of their mathematical models.
So, the idea is sort of not new... Or is it? ;)
The task is to attempt to create a 'Market Function', which we shall conventionally denote by R(t) or simply R. This function will have to satisfactorily describe any regularities on the basis of known functions and their combinations. Anyone can propose his or her own version of the function R, which will be comprehensively tested for strength and reliability. Let us imagine, for instance, that I have my own version of this function whose form I don't want to divulge yet to avoid premature attacks. I would like to disclose what I know at this stage: It adequately describes the exponent, power and exponential functions, the branches of hyperbola and parabola, easily transforms to a straight line, describes any combination of the above functions combined into one function by their addition (subtraction) or multiplication (division), including the sine wave. For example, if the input data are given by a pattern y=a+sint, then R "turns" into this function. We will continue the discussion on this topic if participants show interest in this direction of research, or point out another way to find market patterns.
Let's give it a try, Yusuf. We'll use a piece of some quote chart as inputs. Without revealing the interior of your function R(t), you can demonstrate its output. And the subtleties will appear as we go along.
Dear avtomat, I suggest to check your idea at the final stage of researches, because it is our final goal, and now it is necessary to make sure that R-function proposed by someone, in particular by me, correctly describes artificially created, based on rather complicated, but known regularities, numerical series, similar to the series of quotations.