Thoughts on the random - page 18
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
I willingly believe that with a well constructed adaptive filter it is possible to build a profitable TS.
What happened to the difference between the filter output and the initial quotient? On what basis do we neglect this residual?
.
And it is generally accepted that the residual is much more important than the smoothed curve.
1) the difference doesn't go anywhere -- it's all in the works.
2) What kind of common recognition is this "much more important"????
The residual can be important for error estimation. It's risk.
What risk? Think about what you're talking about. We're talking about signal detection - separating the signal from the signal+noise mixture.
Here's a picture to clarify:
Noise is superimposed on the signal - the price is wiggling around the signal red line.
The decision is made based on the movement of the signal, not the noise. Noise, of course, contributes to it.
But we must understand the roles of signal and noise, their correlation, their weighting.
So what does this have to do with risk?
Risk is another matter. And you don't have to mix everything in one pile.
What's the risk? Think about what you're talking about. We're talking about signal detection - separating the signal from the signal+noise mixture.
Here's a picture to clarify:
Noise is superimposed on the signal -- the price is hovering around the signal red line.
The decision is made based on the movement of the signal, not noise. Noise, of course, contributes to it.
But we must understand the roles of signal and noise, their correlation, their weighting.
So what does this have to do with risk?
Risk is another matter. And don't mix everything in one pile.
And how is this signal line different from EMA or SMA?
How is this signal line different from the EMA or SMA?
I see that any explanation is useless...
Oleg, what you call a signal is such only for your system. Similarly, about the noise. It is a characteristic only of your TS.
When I talk about risk, I mean treating the quote flow as a random process. The returns of this process have some distribution, which affects, for example, the setting of reasonable stop losses.
If you make a mistake with the distribution itself or its parameters when determining the stop-loss, you can lose big time. Niederhoffer didn't believe a Russian default was possible either. But it did happen, even though his model estimated the probability of it happening to be beyond 10 conditional sigmas.
Fat tails are a measure of risk in a way.
Oleg, what you call a signal is such only for your system. Similarly, the noise. It's only a characteristic of your TS.
When I talk about risk, I mean treating the flow of quotes as a random process. The returns of this process have some distribution, which affects, for example, the setting of reasonable stop-losses.
If you make a mistake with the distribution itself or its parameters when determining the stop-loss, you can lose big time. Niederhoffer didn't believe a Russian default was possible either. But it did happen, even though his model estimated the probability of it happening to be beyond 10 notional sigmas.
Fat tails are a measure of risk in a way.
OK. I'll leave it out of the way.
Fat tails are a measure of risk in a way.
Alexey, if since late November 1997 Vasya didn't understand that Russia's default was inevitable, then Vasya is to blame, not the tails.
I underline: it was his fault and I would have kicked him out of his job by early December 1997 at the latest.
Alexey, if from the end of November 1997 Vasya did not understand that the default of Russia is inevitable, then Vasya is to blame, not the tails.
I underline: it was his fault and I would have kicked him out of my job by early December 1997 at the latest.
Exactly.
Oleg, what you call a signal is only that for your system. Similarly about noise. It is a characteristic only of your TS.
When I talk about risk, I mean treating the flow of quotes as a random process. The returns of that process have some distribution, which affects, for example, the setting of reasonable stop losses.
If you make a mistake with the distribution itself or its parameters when determining the stop-loss, you can lose big time. Niederhoffer didn't believe a Russian default was possible either. But it did happen, even though his model estimated the probability of it happening to be beyond 10 conditional sigmas.
Fat tails are a measure of risk in a way.
Thinking of all the DSP supporters on this forum and on other trader forums (and there are a great many of them) I have formed a slogan "DSP with a hickey!"
People do not want to understand that there is no signal in quotes as they understand it, just as there is no noise as they understand it.
There is a deterministic component in kotir (let's not consider SB with demolition) which they confuse with signal. One could agree with them (it's not about terminology after all, it's about money) if the difference between the deterministic component ("signal") and the quotient were stationary (almost constant MO and almost constant variance).
For an automaton:
And about the generally accepted. The first line is ARCH, more fully thick tails, the mathematical model for this is FARIMA (fractional integrability, Hurst is synonymous). This is not only a sea of literature, but also widely available ready-made, free code (R) which takes into account a lot of nuances in the named.
I wish you success, you automaton. I have confidence that you can build a fairly stable system based on a good adaptive filter, and the obligatory stops can keep it from going down, especially if you know exactly how your system reacts to the delta function (see avalsa above).