Machine learning in trading: theory, models, practice and algo-trading - page 1156
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
It is necessary to do bluntly preprocessing of the initial BP of returnees, smoothing outliers, etc. in order to obtain stationarity.
According to Kolmogorov, only returns and cumulative sum of returns in a sliding window can be predicted, as they have expectation=0.
Well, what prevents to convert the price to a fixed increment with a given range, for example n0 = +500 points, n1 = -500 points.
Returns will be the same size of 500 pips, all that's left is the direction.
Yes, yes, we've heard it 100 times ... and tried it 100 years ago, and threw it away because it's useless crap
Did you look at sorcerer's pictures carefully? Did you see what kind of BP he's working with?
What he then does with it - I do not know, but at the beginning it 100% handles a number of returns.
First of all - stop repeating the same thing about returns, Kolmogorovs and stationarity.
You've worked with returnees, so what? You got blown away in the first trend, and more than once.
Secondly, there are a million ways to make the price look stationary, take even a stochastic, why don't you get a stationary series? Or a double or triple differentiation of the series will also produce a super stationary series!
So stop spreading this nonsense about stationarity, I'm sick of it.... It doesn't work, and you know it better than me, but you're broadcasting it like a robot, just like Sanych with his garbage
Enough about returnees and stationary, it's time to change the tune.
Thirdly, enough with the sorcerer.
Did you look at the sorcerer's pictures carefully? Did you see what VR returns he works with?
What he then does with it - I do not know, but at the beginning he 100% handles a number of returns.
I'll tell you more, you do not know and what he does with it before, you can not even think that his series may be synthetic, made up of two or more tools, and it will immediately be + - Static by itself ...
Then why are you whining if you know everything?!
State!
Then why are you whining if you know everything?!
State, come on!
I'm not whining, I'm clearing this thread of nonsense.
Google it, there is a lot of stuff in the same R-ka
Yes, yes, we've heard it 100 times ... and tried it 100 years ago, and threw it away because it's useless crap
Golden words Victor Benedictovich.... Hi there!!!!
ZZ as a Target - facespalm
And so, yes, the constructive
This is the details, they may certainly be the devil, but the participant has shown how and what to do, briefly and to the point, frankly, nothing more sane in this entire thread and the entire forum, saw articles Vladimir Perervenko they are also good, but the essence of the same, only very much smeared and scientific, When they try to put it on the market, it is a sign of ignorance or another author's goals, especially when it is CNN, it makes you laugh and sin.
Knowing in advance that no one will do anything of what I suggest, but still.
1. Everyone knows that the real BP and its increments are 98% like a stationary Laplace motion random process, except for the 2% anomalous outliers in the "heavy" tails.
2. Attaching 2 rows of pure Laplace motion
3. I suggest you try it - can your neural network or forest make money from these rows.
4. If it works, the main task is to convert the original BP to Laplace motion
5. If it doesn't work, then train NS and trade only on those parts of BP, where there is a deviation from Laplace motion by corresponding features.
Knowing in advance that no one will do any of what I suggest, but still.
1. Everyone knows that the real BP and its increments are 98% like a stationary Laplace motion random process, except for the 2% anomalous outliers in the "heavy" tails.
2. Attaching 2 rows of pure Laplace motion
3. I suggest you try it - can your neural network or forest make money from these rows.
4. If it works, the main task is to convert the original BP to Laplace motion
5. If it doesn't work, you should train NS and trade only on those parts of BP, where the deviation from Laplace motion by corresponding features is observed.
I'm not dealing with the gradients, but I downloaded the file, and there are some figures in one column - and what to do with them, how to generate the predictors?
When I downloaded it, I thought that it had at least target ones and the history is suitable for loading in the terminal, I personally generate predictors there.
I do not deal with increments, but I downloaded the file, and there are some numbers in one column - and what to do with it, how to generate predictors?
When I downloaded it, I thought that there are at least target ones, and the history is suitable for loading into the terminal, I personally generate predictors there.
It is an artificial series like Laplace motion with a start at 0 and nothing more.
So far my TS is giving conditional +175.86 pips on the 1st series of 25 trades (+14/-11). I'll check it further...
Does the NS get anything out of them?