Recognise changes in the "behaviour" of a financial time series (Trading on the news) - page 4
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
geez... It's not the accuracy of the fit, it's the accuracy of the prediction itself. We limit it, in real time. We say the model will be considered relevant to the current situation if its forecast falls within a certain narrow range. If not, we consider this point to be a "breaking point" and either put another, pre-prepared model into effect, or, for lack of one, we simply minimise it and wait.
And I've been checking this out. And out of sample and of all sorts. It is of no use. I don't know a method which would tell me if a model adjusted (adapted!) to a newly arrived candle is good for this new sample - you can tell about it when this new candle with the kink becomes not very right, but moves some distance deeper into the sample. If there wasn't a kink, then I could make models with a profit factor higher than 10.
Read the attachment.
I don't know the method
Doesn't mean it doesn't exist. Who says that AR-MA model is the only way to encode the behaviour of a series on the fitting interval? There is, for example, fractal compression, which seems to work better here.
It does not mean that it does not exist. Who says that the AR-MA model is the only way to code the behavior of the series on the fitting interval? There is, for example, fractal compression, which seems to me to be more appropriate here.
I haven't used ARMA, but that's not the point.
Different models give different breakpoints, which means that there are worse or better models with respect to breakpoints, not forecast accuracy. Finding a "good" model in relation to breakpoints I've put off for now, as there's a lot of potential in being able to use the forecast. A simple example. Supposed we have the last price 1.3000 for EURUSD, the forecast is 1.3025. The question is: should we enter long or short? So, if I penetrate deeply, the answer, though obvious, is not obvious. We need to consider a number of additional factors for an answer.
It depends what you mean by this. The beginning of a new trend, the end of the old one, and stagnation? Or maybe you want to know that Citibank decided today to balance their portfolio a bit?
But in general, this is a problem about the so-called process disintegration. cf. Kolmogorov, Shiryaev, etc. A.Gorchakov, who is not unknown in trading circles, even wrote some formulas about it on his website.
In my opinion, this is all nonsense. Not bad for the task of detecting a target amid radio interference, but not for the market. However, no one can stop people from trying to pull the market on formulas.
Are there no break points outside the news?
Of course there is. The market is moved by news, but this is not necessarily visible to the eye.
What is a kink? In order.
We have a TS. Initially, if it is a TS, it corresponds to the market and correctly identifies the entry-exit. Then we have to optimize it by adjusting some parameters of TS according to changes of the market. If it works, then the relationship between the parameters and changes in the market is obvious. But it is on the surface. If the TS is made on the ideas of econometrics, it can be seen that TS parameters are random numbers, i.e. non-deterministic quantities and in fact we are dealing with estimation, not quantities. And since this is the case, the question is about the law of distribution of this SV. It's about the parameters. But the market can change in such a way that the functional form of the TS needs to change. I.e. when I talk about kinks, I'm talking about kinks in the models on which the TS is built. So you can't tell anything from the kotir graph. By the way, I didn't give a complete classification of the breakpoint types.
However, no one can stop people from trying to pull the market on formulas.
Of course there is. The market is moved by news, but this is not necessarily visible to the eye.
What is a kink? In order.
We have a TS. Initially, if it is a TS, it corresponds to the market and correctly identifies the entry-exit. Then we have to optimize it by adjusting some parameters of TS according to changes of the market. If it works, then the relationship between the parameters and changes in the market is obvious. But it is on the surface. If the TS is made on the ideas of econometrics, it can be seen that TS parameters are random numbers, i.e. non-deterministic quantities and in fact we are dealing with estimation, not quantities. And since this is the case, the question is about the law of distribution of this SV. It's about the parameters. But the market can change in such a way that the functional form of the TS needs to change. I.e. when I talk about kinks, I'm talking about kinks in the models on which the TS is built. So you can't tell anything from the kotir graph. By the way, I haven't given a complete classification of fracture types.
I was exaggerating)
Any TS uses formulas, because all your indicators and other nerd stuff have formulas implemented in the form of programs. Therefore you either unconsciously "stretch" the formulas, or deliberately, and even after reading books, and do not read different Shiryaevs.
It's not about formulas in general, it's about a specific mathematical apparatus. It's all nerdish for nerdish's sake.
The second point. There are strategies for trading on the news that are surprisingly simple and old.