Econometrics: one step ahead forecast - page 66

 
Mathemat:



Cool.

So, I go to the market in Tula province and say: Grandma, how much is the muv for a price? And she says: Go away, you wretched man, the muv is not for sale at any price, you impertinent... ...and smiles carnivorous like Peter. Where is he, by the way?

 

What the fuck?

I type one thing, and another appears on the forum, not mine...

 

Круто.

That's what I'm saying: it's all too sweet. The reality, of course, is tougher. We don't know the initial moment of accumulation. I have tested this in different situations. Sometimes it comes out that you need to accompany a sold muv for another dozen steps (on muv-13) to wait for at least zero difference between price and muv (crossover).

But if your deposit is $100K and you are working with 0.1 steps at accumulation, you will have good luck with small drawdowns of equity. The trend will not last forever, it's an axiom.

 
faa1947:

Trendiness is certainly a curious property.

But we should all be interested in the predictability of the quote at the point of the last bar. What can be taken from the history to make a conclusion about predictability? So far I can't add anything to what I've laid out here. These are the following requirements for the model:

Try Fuzzy Logic. It may be more suitable for such tasks than a neural network.
 

Once again, I'm posting a picture so you don't have to go to the previous page.

I am cheating you here and you are falling for it. Exactly in this situation, it was possible not to redeem the price, as it was already going down.

But the reality, I repeat, is tougher. It is not always this beautiful, because we do not know where the price will be in 13 bars. It is possible to start the "Y" operation at any point in time. But sometimes we will have to wait for a long time, very long, up to 100 bars or even more.

If we choose an unsuccessful moment to start the operation, we will have to follow the already accumulated muv all the time to wait for the successful moment of price redemption. And then we will have to wait for the moment when the whole operation is completed.

 

Colleague, I did not see here SanSanych's attempts to promote neural networks, nor the prospects of using fuzzy sets Via regression analysis.

What I have seen is an infrequent combination of professionalism and determination, and therefore I expect interesting discussions and results that are not expected nowadays.

 
Mathemat:

The muv equals the arithmetic average of the prices on the different bars. In the initial stage of muv accumulation, we sell the price :) After 13 steps, it turns out that we have sold a muv. More precisely, the sum of prices is sold. But at the 13th step we buy the price in an amount exceeding the steps of accumulation exactly 13 times.

I just took a simple machine. With the others will be much more complicated.


What if, while the muve was being accumulated (sale), the price went above it?
 
Avals:

What if while we bought back the muv, the price went above it?

Did you buy back the muv or sell it?

I see I've corrected myself.

And if while accumulated muv (sale), the price went above it?

We continue to support the muv, by covering the first sold portions and selling new ones. We are waiting for the moment when the price is significantly below it again (this is the condition, in which we finally enter the stock). As soon as that happens (the price may be higher than the original price, by the way!), we immediately buy the price.

The first part of the operation is completed. Now we wait for the price to rise above it again. And then we will close all positions.

I am not speaking in the language of netting now, but in the language of DC accounting.

 
Mathemat:
Did you buy back the muv or sell it?


sold - while the Y operation was going on
 
DhP:
Try Fuzzy Logic. It may be more suitable for solving similar problems than a neural network.

I don't use neural networks and I don't believe in them at all.

I don't go down the route of "let's try another gimmick". I identify the problem and try to solve it. Pretty much solve it myself in the hope that someone will join in, but it's not happening.

I articulated the criteria for "goodness" of the model above, but that's not interesting. I'm waiting for it to get interesting.