Discussion of article "Universal Regression Model for Market Price Prediction" - page 2

 
yosuf:
k is the sample size each researcher chooses for himself and I don't understand the question t>k am I limiting k? take as much of it as you want.

Is k the sample size of the actual price values? Or is it something else?

It seems so to me from your explanations of the final form of the regression equation. Here's what you write:

∑Pf= P0+ P1 + P2 + ...+ Pk is the sum of the actual price values;

 
yosuf:
Why am I obliged to post exel? If you mean my offer on the forum about searching for developers to translate exel to mql4 for MT4, then it was meant that further conversation will go on a confidential level, and this offer remains in force, especially I made a statement known to you today at 9.00 Moscow time.
if you don't want it to be publicly known, then please send it to me in private.
 

Yusuf,

Теперь допустим, что скорость V(t) изменения рыночной цены P(t) пропорциональна как величине D(t), так и времени t:


Why is it proportional to time? If we use common sense, it is rather inversely proportional: the price due to some influence changes slower the further in time we move away from this influence. Although, by and large, it is not shown to what extent the time dependence on V(t) is already accounted for in D(t). So I have to consider your statement unsubstantiated and, moreover, in all probability, incorrect.
 

D(t) + H(t) + P(t) = D0                                                                                                                                     (4)


Yusuf, you have the first and third summands in this equation have the dimension of price, but the second summand has the dimension of the rate of price change: although you have defined in words that H is only numerically equal to V, you have not added the multiplication by the unit of time in the corresponding formula (which, as we know, can be arbitrary). Hence the logical (and mathematical!) error: the underwritten unit of time violates the dimensionality of equation (4) and makes it meaningless.

Actually, it looks very much like you just picked up some functions that give in sum some "material balance" by the method of not quite scientific poke, and then you are engaged in ordinary self-deception. I do not see a single argument in favour of choosing one or another type of dependencies P(t), V(t) in the article, while you essentially base all your reasoning on the fulfilment of these dependencies.

 
Vita:

Where P(0)corr and Dcorr contain information about the sum of the actual price values and the trend direction of the actual data for t = 0,1,2,......k;

Does your formula assume forecasting for t > k?

I assume that you have not compared your regression forecast in any way with the linear regression forecast for t > k?

1. Look at equations (15) and (6), find the answer to your first question, again you don't understand, refer.

2. Assumes for all t to infinity

3. Why. The regularities of market price do not even smell of linearity

 
Vita:
According to the legend of the above graph, P1 is a dark blue smooth curve. Is it it somewhere from the 38th count showing the prediction?

1. Up to the 38th point the robot is engaged in the analysis of facts and their approximation by formulas (11)-(14), and then it starts to evaluate the situation, formulate verdicts regarding its further behaviour on the market and determine the trading strategy by formulas (15)-(19), by the way, it does this starting from the third point continuously, as information from the market arrives, even in the form of ticks, it manages to do all this before the next tick, no matter how short or long they were. With two points - the robot is blind and helpless, but it does not harm its master, it does not even analyse the market, the robot and the whole trading process is triggered by the third point or tick, as you wish, and from the fourth point it is ready to launch an avalanche of orders in the necessary direction, if the situation on the market and the condition of justifying the spread in the multiplicity that is set by the master allows it

2. This seemingly calm at the moment, appraisal-forecasting line, to which you point, is a trace of the moment when the robot with all its fantastic power collapsed with an avalanche of sell orders and calmly earns without seeing any fear in the foreseeable future, but at any moment, which he himself determines, he immediately closes all orders, at that the former calm curve acquires a terrible look, bifurcates, on the left side becomes like a tail of an avalanche of orders.similar to the tail of a swallow, on the right to the rails, indicating the ticks of future possible amplitude maximum and minimum values of prices with a stick thrown across the rails, indicating a possible moment of trend change and constantly moving it along the rails of price ticks indicating changes in the moment of trend change according to formulas (11)-(19), as the forum participants yesterday during the testing of the robot to create an expert for mt4 on the programme made in exel, even one of the participants asked: this terrible kind of curve you call forecasting, nonsense, and I postponed for later the announcement of such a situation.

создать эксперт для мт4 по программе, выполненной на экзель - MQL4 форум
  • www.mql5.com
создать эксперт для мт4 по программе, выполненной на экзель - MQL4 форум
 
Vita:

k is the sample size of actual price values? Or is it something else?

It seems to me so from your explanations of the final form of the regression equation. Here's what you write:

∑Pf= P0+ P1 + P2 + ...+ Pk is the sum of the actual price values;

YOU are right
 
alsu:

Yusuf,

Why is it proportional to time? If we use common sense, it is rather inversely proportional: the price due to some influence changes slower the further in time we move away from this influence. Although, by and large, it is not shown to what extent the time dependence on V(t) is already accounted for in D(t). So I am forced to consider your assertion unfounded and, moreover, in all likelihood incorrect.
You didn't read the article or didn't pay attention to it, where it is approximately as follows: since imperatives (4), (5), (9) and the balance (19) turned out to be fair, we conclude that all the assumptions made are fair, I recommend you to read the article more carefully again in order to better clarify this situation for yourself
 

In the list of used literature there is nothing at all on the subject of transients. In connection with this question to the author - how much are you familiar with the theory of transients? It seems that there is a process of new discovery (or new development) of the theory of transients, but it has been discovered and worked out long ago. The question remains - what to do with this theory, or what can be taken from it? The picture of the transient process is a sum of exponentials, and the exact picture depends on the parameters of the medium and the magnitude of the influencing impulse. We do not know the characteristics of the medium, we do not know the strength of the impulse and the moment of time of its occurrence. In essence, the problem is reduced to exponential regression (or rather to logarithmic regression). But from where and to where to approximate, and from where and to where to extrapolate?

 
alsu:

Yusuf, you have the first and third summands in this equation have the dimension of price, but the second summand has the dimension of the rate of price change: although you have defined in words that H is only numerically equal to V, you have not added the multiplication by the unit of time in the corresponding formula (which, as we know, can be arbitrary). Hence the logical (and mathematical!) error: the underwritten unit of time violates the dimensionality of equation (4) and makes it meaningless.

Actually, it looks very much like you just picked up some functions that give in sum some "material balance" by the method of not quite scientific poke, and then you are engaged in ordinary self-deception. I do not see a single argument in favour of choosing one or another type of dependencies P(t), V(t) in the article, while you essentially base all your reasoning on the fulfilment of these dependencies.

Good right question with a hasty wrong, in my opinion, auto-answer

1.According to my newly proposed vision of dynamic transient processes, the philosophy of the process, I divided the whole cycle from the moment of price destabilisation to its completion into three periods-future (L), present (M) and past periods. It turns out that in this process the integral unitary function (L) is the primordial one, as can be seen from (1) and (6). How does its formation take place and why is the market driven by the future? How to understand this fact? There seems to be a contradiction with our ideas about the causes of market destabilisation, it seemed to be related to the past. I decided to understand this inconsistency in the following way: before price destabilisation, first tensions build up, the market tries to stabilise the market by the law of supply and demand, but the residual unhappy tensions accumulate into a unit function (L) to hit the market at a certain point in time ,taken by us as t=0 with the first stabilising, from the point of view of the market, (and from our point of view - destabilising) portion, increasing the price by an amount proportional to the non-unitary function (M1), having the maximum possible value 1/e = 0.3678... The next portion (L) pushes the first portion (L), which was numerically equal to (M1), into the past, which gradually add up in the past (i.e. all pushes back portions (M)) to form a unitary (because all unitary (L) eventually passes into the past through the present (M)) integral (because all (M) coming from the present are summed up) function (S), which will eventually absorb all (L) through (M). The new price is fully formed on the market, the tension is removed, from the point of view of the market and it should seemingly calm down. Mathematically this process is shown by two formulas without numbers between formulas (8) and (9). the continuation of the answer will follow later.

2. Now the question asked by the participant about "non-conformity of dimensions" has been solved by itself: Indeed, we see that the functions L and S need not have any dimensionality, since they are integral and unit and/or zero in the corresponding opposite boundary conditions, so, thereforethe functions D(t) and P(t) have their dimension of prices inherited from D0 according to the above relations, and with the function H(t) it is a little more complicated, since both its components, both the differential function M and D0have dimensions, giving by product the dimension price divided by time, and it should enter into the tandem of functions D(t) and P(t) exactly with such dimension to form D0 in sum, but it has only one chance to realise it - to temporarily "hide" its second dimension "time" and numerically show its value with the dimension "price", therefore in the article I had to limit myself to a dry statement of this fact, which you have taken for inconsistency. This metamorphosis is easier to explain by a simple example: let's imagine that the price of currency in a few days should increase by D0=10 roubles. For the first day of the price increase, let's say, the price increased by 1 ruble, then the balance looks like this: D(t)+H(t) +S(t)=D(1)+H(1)+S(1)=9+1+0=10=D0 The next day the price increased by 3 rubles: D(2)+H(2)+S(2)=6+3+1=10=D0. In principle, it would be possible to write D(t)ruble + H(t)ruble/1 time*1time +P(t)ruble=B0 ruble, where the dimensions are respected, but I pedpochlal to write as written, which is the same.