Random Flow Theory and FOREX - page 27

New comment
 
Thanks, Rosh, but I haven't seen that link. Let's see what kind of beast it is...
 

Once again I don't understand what's going on in my head, so I'll try to explain

What happens when you look at the market (war).

  1. I shoot, hit, 1 did not hit 0.
  2. The enemy shoots, hit 1, did not hit 0
    3. .

This is not a blame game where everything is mixed between the enemy (the market) and myself (the trading system). Yes it is necessary to analyze the joint hostilities, and at the same time it is necessary to separately study the enemy, and my behavior. That's exactly the point I wanted to make to mech. mates, in the H-volatility thread. It's just that after the mathematician's link, all the bricks in my head have added up.

With this approach the flies are separate, the cutlets are separate. And not all in a pile and it is not clear what to do with it. Simply then a completely different theory comes into play, I would call it the theory of "war" games. After all, the analogy is direct, just think about it. My opponent shoots me 100 times and hits me 20 times (he is easily wounded), I crawl over and kill him with 1 precise shot to the head. This is very similar to sitting on the flat, suffering small losses and if not killed, was able to crawl to take profits on the trend. If I cannot kill them, I cannot crawl back. I may vice versa, I shoot my enemy a bit, wound him all the time, the main thing is that he does not kill me with 1 shot, but lets me live as long as possible till he dies bleeding at my feet in the form of AC refusing to take profit (I think you will find the analogues yourselves :-)).

If you look at the market from this angle, at least for me, it helps a lot. I began to see the goal of research more clearly and the way to achieve it. It is now not just a matter of building a good TS (my weapon) that is good at engaging the enemy, but also a system of defense and combat tactics. Taking the most advantageous position in a battle, when the probability of hitting me is much lower (preferably infinitesimally small) than the probability of hitting me.

So that's it.

Z.U. I suggest before it's too late to get on my side of the monitor :-), for 1 in the field is not a warrior.

 

The combat tactics presented above are extreme points, I wouldn't want to work in them. There's a lot of interesting stuff in the middle there, from sniper shot from cover, to shooting on the pass. Eh, even a grenade could be thrown there (kind of like how the grandfathers from Mechigan's published the figure, the market and rushed to another shelter), and we with the large-caliber on the pass :-). That's where the notion of an insider got in :-)

 
Prival:

What do you get when you look at the market (war) in this way?

  1. I shoot, hit, 1 did not hit 0.
  2. The enemy shoots, hit 1, did not hit 0
    3. .


Sergei, don't pay attention to anyone. For a man to be able to set himself a task so that he can solve it, it is not enough to do it correctly from, say, a mathematical point of view. There are a lot of other aspects on which this depends to no lesser extent. For example, the psychological aspect or the aspect of visual perception. Without them - nowhere.

If I do not understand the dry language of mathematical abstractions, then no matter how I formulate the problem in it, I still will not solve it. But if I find the situation in the real world, where the same laws are manifested, then my chances are immeasurably higher, because I begin to perceive it through my physical way of thinking. And if your perception is focused on the task of hitting a moving target, then you should not walk away from it, but use it.

I would just make some adjustments. The objective is to hit the enemy aircraft. Aggravating circumstances - the classical laws of physics do not apply, and which ones do should be investigated and established. The weapon is an unguided rocket with a limited range. The only option for controlling the launched rocket is self-destruction. If it happens within range - profit, if outside - loss. The main condition is that the number of missiles is limited, and replenishment depends on the success of firing.

And the psychological aspect is simple - if you shoot all the missiles and the plane flies away, the family will have nothing to eat.

 
Yurixx:
Prival:

What do you get when you look at the market (war) in this way?

  1. I shoot, hit, 1 did not hit 0.
  2. The enemy shoots, hit 1, did not hit 0
    3. .


Sergei, don't pay attention to anyone. For a person to be able to set himself a problem so that it is solved, it is not enough to do it correctly from, say, a mathematical point of view. There are a lot of other aspects, on which it depends just as much. For example, the psychological aspect or the aspect of visual perception. Without them - nowhere.

If I do not understand the dry language of mathematical abstractions, then no matter how I formulate the problem in it, I still will not solve it. But if I find the situation in the real world, where the same laws are manifested, then my chances are immeasurably higher, because I begin to perceive it through my physical way of thinking. And if your perception is focused on the task of hitting a moving target, then you should not walk away from it, but use it.

I would just make some adjustments. The objective is to hit the enemy aircraft. Aggravating circumstances - the classical laws of physics do not apply, and which ones do should be investigated and established. The weapon is an unguided rocket with a limited range. The only option for controlling the launched rocket is self-destruction. If it happens within range - profit, if outside - loss. The main condition is that the number of missiles is limited, and replenishment depends on the success of firing.

And the psychological aspect is simple - if you shoot all the missiles and the plane flies away, the family will have nothing to eat.


Then only a sword will do the trick. It's not about swinging a hatchet, though.
 
lna01:
Prival:

Candid I have a request if it is not difficult to check ACF fig.3, if it is the same then there is no point in checking acceleration, if so the SRS system will consist of 2 equations.

First question: why did you take returns for the original series and not for Y-mu?


You're right not to take returns in pure form :-(, it kills the trend. I can't restore the initial process back. I never really thought about it, I thought it was always possible to return to an exact constant.
 
Prival: You're right not to take returns in their pure form :-(, it kills the trend. You can't restore the original process back. I didn't even think about it, I thought it was always possible to return to an exact constant
.
So it does, so what? Prival, trend killing plays the most beneficial role in this case by removing false autocorrelation related to the trend and making counts more independent. In the original series the best prediction is, roughly speaking, the previous value (I won't throw in the "m" word, which you decided to discard), i.e. the counts are clearly dependent. In detrended by taking the first difference we achieve the elimination of most of this dependence.

And who said that the first difference series (returns) cannot be used to restore the original series, having the value of the original (prices) at least at one point? It is the same "discrete" derivative by which the original function is reconstructed.
 
Mathemat:
Prival: You're right not to take returns in their pure form :-(, they kill the trend. I never gave it much thought, I thought it was always possible to return to an exact constant
So it does, so what? Prival, killing the trend in this case plays the most beneficial role by removing false autocorrelation related to the trend and making counts more independent. In the original series the best prediction is, roughly speaking, the previous value (I won't throw in the "m" word, which you decided to discard), i.e. the counts are clearly dependent. In detrended by taking the first difference we achieve the elimination of most of this dependence.

And who says that the first difference (returns) series cannot be used to restore the original series, having the value of the original (prices) at least at one point? It is the same "discrete" derivative by which the original function is restored.


I guess we don't understand each other again. Forgive the military fool, that's not how my mind works. Let's check your statement "And who says that from a series of first differences (returns) one cannot reconstruct the original series with the value of the original (prices) at least at one point? It's the same "discrete" derivative from which the original function is reconstructed. "

Methodology.

  1. Remember Close[0] is a constant, which is necessary for reconstruction of the original series
  2. Take the transformation Close[i]-Close[i+1].
  3. Take the reverse action Close[i]+Close[i+1].
  4. Add constant Close[0].
  5. We compare the original row of numbers on the chart Y[i] with the one obtained using the inverse transformation on the chart YYY[i] (fig. 1).

Fig. 1

As you can see, the red curve (the original row of numbers) does not coincide with the green curve (that is, what you get after the transformation). The total error = 746 points.

Now let's take another method (sequence of actions).

The same as in the first methodology, the only difference is that we take into account the trend, in this case I understand the straight line equation y(x)=a*x+b in Fig.

I.e. we don't perform transformation with Y[i] right away, we preliminarily subtract mu, and of course it should be taken into account again when performing reverse transformation. Here we have Fig.2

Initial curve is completely restored Total error =0. Thus, I claim that

  1. This transformation (returns) kills the "trend"
  2. The original series cannot be restored by inverse transform
  3. The cumulative error is directly proportional to the sampling depth and the coefficienta in the straight line equation.

Please double-check this statement as either one of us is wrong, either Mathematician or me.

Or we again get confused in terms and do not understand each other.

 
Prival:
Mathemat:
And who says that a series of first differences (returns) cannot restore the original series, having the value of the original (prices) at least at one point? It is the same "discrete" derivative by which the original function is restored.


I guess we don't understand each other again. Forgive the military fool, that's not how my mind works. Let's check your statement "And who says that from a series of first differences (returns) one cannot reconstruct the original series with the value of the original (prices) at least at one point? It's the same "discrete" derivative from which the original function is reconstructed. "

Methodology.

  1. Remember Close[0] is a constant required for reconstruction of the initial series
  2. Take the transformation Close[i]-Close[i+1]
  3. Take the inverse action Close[i]+Close[i+1]
  4. Add constant Close[0]
  5. Compare the initial series of numbers on the chart Y[i] with the one obtained in the inverse transformation on the chart YYY[i] Fig. 1

Thus, I argue that

  1. This transformation (returns) kills the "trend"
  2. It is impossible to restore the original series by the inverse transformation
  3. The total error is directly proportional to the sampling depth and the coefficienta
    4. in the straight line equation.

Please double-check this statement as either one of us is wrong, either Mathematician or me.

Either we are again confused in terms and do not understand each other.

Returns cannot "kill" a trend! And of course it is possible to restore the original series to a constant by simply summing up the residuals - this is the derivative-integral transformation. Let's do everything according to your methodology, see above:

The first figure shows the original and the restored series. The second one shows their difference. Where is the effect of "trend destruction" (counts 500-700)? There is another problem here. The detrending operation (when smoothed BP is subtracted from original BP) adds to difference series dependences that do not exist in original series (imaginary correlation). This has to be kept in mind.

Thus, I argue that

  1. This transformation (returns) does NOT kill the "trend"
  2. Restore the original series by inverse transformation MAY
  3. result in a total error = 0.

    So Mathemat is right .
 

Neutron

Please select an area where there is a trend. In your case (it is not visible in the sample). Plot the trend curve on 1 graph. You have coefficient a=0. The other sample please.

1...202122232425262728293031323334...85
New comment