Bayesian regression - Has anyone made an EA using this algorithm? - page 36
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
Ladies and gentlemen, ladies and gentlemen, comrades! There's too much blood in your alcohol system.
What can be mathematically modelled on R if you haven't decided on the conceptual issues for Bayesian formula: what is the market to the right of the zero bar. And is it a market? Or maybe a good game simulator with an appropriate algorithm?
Of course there's no blood in our port. But it makes absolutely no difference who is behind the market on the right. Statistics are enough. Let's remember Wiener and anti-aircraft fire control system, his book - "Cybernetics 1948", where it is described, hopefully, on your shelf, it is not a problem to find it in the Internet too.
The movements of the aircraft are far from random, but to you they are completely chaotic. Still, anti-aircraft fire may be quite effective.
As for game simulators, there are at least a few and quite independent, and this is already close to a normal distribution.
According to research by a mathematician (I don't remember his last name, he works for FINAM), the distribution is close to normal with elongated tails (but it's understandable why). So linear regression, imho, quite rules.
Let's remember Wiener and the anti-aircraft fire control system, his book - Cybernetics 1948, where it is described, I hope you have it on your shelf.
You think well of me.
As for game simulators, there are at least a few and quite independent, and that's close to a normal distribution.
Do you also think forex is more of a game simulator than a market?
Isn't he the one who wrote that he held a position in some analytical-mathematical department of the KGB and asked, I believe, 49,000 rubles for several seminars? ?
It's probably him. I read a report at a conference, and I attended a couple of his seminars a few years ago.
Actually, there were some strong mathematicians there... in their field, I had to talk to them.
Do you also think that forex is more of a game simulator than a market?
I accept that. As do the domestic markets, for that matter. Some points I know for a fact, voiced by the same FINAM and IT Invest.
There are market makers.
Maybe look at how noisy the market is? After all, it is no secret that there are markets with a lot of noise and ones with less.
The more noise, the less trend-following strategies will work.
I accept that. As do the domestic markets, for that matter. Some points I know for a fact, voiced by FINAM and IT Invest.
There are market makers.
The normality of the distribution can only be established on the general population. If we do not have a general population and we do not have any others, then we have to prove that the mean tends asymptotically to its mathematical expectation. And if this result is obtained, characteristics obtained on any limited segment of the general population can be freely extrapolated to any other segment of this general population.
We have nothing of the kind on the market. All the figures of R2 given above are nonsense, as there's no proof that the selected plot is a part of the general population that has at least the stationarity property. Therefore the figures were obtained on the specified areas, but they have nothing to do with the future: they may coincide, they may not, they may coincide 100 times and then they sell the deposit along with the profit.
This is why the whole world is so obsessed with the stationarity of the raw data - it is the rationale for extrapolating statistical characteristics obtained in one area to other areas. This is why "out-of-sample" calculations are made, and worst of all, if the values in one plot differ from those in another plot.
Increasing the size of the training sample does nothing. The Eurodollar started with 1, then 0.9, then 1.6 - will you sit through 15 years waiting for the right movement?
To build a TS you need some reasonable window + considerations that the characteristics of that window can be extrapolated into the future.
And lastly, what is applicable?
It depends on what and for what. The choice of target variable is a matter of principle
There are actually two directions in regressions:
I am in favour of classification.
Have you thought about which data (over which time period) to use to calculate the regression?
Should the data be static or variable depending on the market pattern?
"The original intent was to combine a straight line and a price series." - if the Bayesian regression is a straight line, then it is really no good.
You don't need a straight line.
If the price series is represented as:
C = analytical function + noise, then the noise will be normally distributed, imho.
The price series itself is more of a Wiener random process - random walks.
ZY analytical function, e.g. Fourier series.