Econometrics: one step ahead forecast - page 79

New comment
 
That's not what I meant at all :)) Oh, come on.
 
Reshetov:
I understand, so I apply a safety margin.

Why not a triple? :)
 
Avals:
why not a triple? :)
Bojan :) already mentioned :))
 
TheXpert:
Bojan :) already mentioned :))

things change fast, I can't adapt in time :)
 
Reshetov:

No conclusions can be drawn.

You take the vehicle and start driving, then do the same amount or 10 more of the same amount. But you are not interested in what kind of vehicle you have: whether it is a trough that goes downhill, or a tractor, or a Mercedes.

I took econometrics (I knew statistics long ago) not for the good of life.

I have no problem to write a TS with a profit factor over 5. I test everything scientifically. I start trading. I am optimizing, then again everything is fine, but then profit factor starts falling down again. And in the end I have to withdraw my TS completely, as it cannot be adjusted. The most annoying thing is that I cannot distinguish between drawdown and drawdown.

So the question is: is it impossible to judge the future of the TS one bar ahead (which TS are we riding? based on the analogy above) based on the internal structure of the TS? That doesn't rule out everything you write about testing. But here's to forming an opinion before testing. With my one step forward approach followed by adaptation, any model is derived to profitability by MM. But what is wrong with it when the profit factor in observations is less than one in the one that is laid out? In TAP they can calculate not just the forecast, but the forecast horizon as well.

 
faa1947: There is art and there is science.

Then econometrics as you have it is not science either. And what kind of science can there be in an arbitrary pile-up of tests, even the strictest statistical ones? And why do you limit what is scientific to the arsenal of what EViews contains?

Econometrics cannot provide sufficient predictive criteria. Necessary all you want, but that doesn't get us any closer to solving the basic problem.

The problem is that you, having blindly believed in these tests, do not even try to go beyond EViews to understand why they do not provide guarantees.

 
faa1947:

I have no problem writing a TS with a profit factor over 5. I test everything scientifically. I start trading. At first all is fine, and then the profit factor starts to fall, I optimize it, again all is fine, and then again the profit factor falls down. I have to finally pull out the TS, as it cannot be adjusted. The most annoying thing is that I cannot distinguish between drawdown and drawdown.

The most annoying thing is that I cannot distinguish between drawdown and drawdown. And if it is 5, it means that it can be thrown away immediately without optimisation.
 
Beware of random scientific connections )
 
faa1947:


You take the vehicle and start driving, then do the same amount or 10 more of the same amount. But you don't ask what kind of vehicle you have: a trough that goes downhill, or a tractor, or a Mercedes.

Interested, but outside the sample, i.e. on a real road. And what econometrics offers is just a fitting for a vehicle - a dummy that looks like a real car in the showroom (on the optimization sample), but when you go to the highway (outside the sample) it turns out to be a trough unsuitable for driving - the wheels fall off, instead of the engine bricks.
 
TheXpert:
I mean something else :)) But never mind.

It's all the same: traders are only interested in the right side of the yield curve - OOS, econometricians are only interested in the left side - the fit.

The bottom line is that neither traders nor econometrists are concerned with the results which are on the part they are not interested in. Therefore traders and econometricians have nothing in common.

1...727374757677787980818283848586...139
New comment