Machine learning in trading: theory, models, practice and algo-trading - page 1642
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
Probably, as you wrote earlier, you can do with the standard means of the MT tester. To be honest, I do not see in neural networks, in and of themselves, anything particularly great. I guess it's not worth avoiding the opportunity to do without them)
Thank you, I guess I wanted some kind of miracle, but you put my flight of fancy down to earth
This is the main problem of evaluation (research) of time series - there is no methodology that guarantees a reliable future forecast of BP (or my estimation of TC portfolio) .... There is no future, is there? - There is only the present and history, all the rest is from the realm of popular movies
((((
This is your target function (just like in AMO), what you want to get as a result of filtering, want to remove noise ? describe your ideal signal and feed it as a reference , want to describe a trend ? same thing.
want to know the "ideal zigzag parameters" at the moment ? describe what the "ideal zigzag parameters" are for you then
try to get "IPZ" on each candle, I think you'll be interested to see it:)
And then you can even try to predict the half-scientific series of "ZPI" by the same ill-fated AMO ))
As a result you get an adaptive system with adequate zigzag parameters and one step ahead forecast, this is what poorIgor Makanu has been looking for a year and still can't find)) at least he's got it written in front of his nose. DearIgor Makanu this is also a solution to your problem "when the system stops working", you can simply monitor the AMO error (based on the zzz parameters) in real time, this will be your criterion of system operability
As a result you get an adaptive system with adequate zigzag parameters with a prediction for a step forward, this is what poorIgor Makanu is looking for a year and can not find)) although they write about it in front of his nose. Also dearIgor Makanu is the solution to your problem "when the system stops working", you can just track the error AMO (the parameters of the buzzer, etc.) in real time, it will be your criterion of the efficiency of the system
alas, it does not work, I searched a long time ago has determined the direction of research, it seems Y.Reshetov and wrote that the NS work in real time as if, but combinations of simple systems obtained by simple methods (without MO) are more effective, but it is worth trying to determine the probability of NS their work in the future
Ok, not the point, once again I get convinced that MO is more a trendy mainstream, on tensorflow I saw off.packages for C# - I wanted to try them, but I don't think so far
Nothing is perfect.
Mmda ... I was expecting a deeper thought, disappointed ...
Alas, it does not work, I searched a long time ago defined a direction of researches, it seems J.Reshetov and wrote that NS work in real time as if, but combinations of simple systems received by simple ways (without MO) are more effective, but it is necessary to try to define by probability NS their work in the future
Ok, never mind, once again I get convinced that MO is more a trendy mainstream, I've seen packages for C# on tensorflow - I wanted to try them, but I don't think so far
Can't you people read yet? )) What's wrong with you?
I said to keep track of the error in the real time.
Here, in general, the main problem of time series assessment (research) is that there is no and will never be any methodology guaranteeing a reliable future forecast of BP (or my evaluation of the TC portfolio) .... There is no future, is there? - There is only the present and history, all the rest is from the realm of popular movies
A good (simple) model of all this seems to me a game model related to the El Farol bar problem. There, too, all there is to the decision to go to the bar or not is just the story. But the rightness of our decision depends not on the story at all, but only on what decisions most other bar-goers will make in the present. The only connection with history is that other patrons may also rely on it to make their decisions in the present and we may somehow try to predict what most of them will decide.
Can't you people read? )) What's wrong with you?
I said error tracking in real time.
It doesn't work!
There will always be an error in such a trading system, and the value of correct forecasts will be determined not by the current deviation of the trained model and the current price (even though you generally mean ZigZag), but by the number of series of continuous losses and profits
Do you realize that it is important to achieve a prediction of a series of trades? - At any moment you can open sell and buy orders, and it will be a correct prediction! But not every order can be closed with a profit
UPD: on a price series you can ALWAYS build an alternative ZigZag and its tops will not coincide with your ZZ - and that ZZ will be correct too do you understand?
A good (simple) model of all this seems to me to be the game model associated with the El Farol bar problem. There, too, all there is to the decision to go to the bar or not is just a story. But the correctness of our decision depends not on the story at all, but only on what decisions most other bar-goers will make in the present. The only connection with history is that other patrons can also rely only on it to make their decisions in the present and you can try to somehow predict the decision of the majority of them.
Game theory has watched and read a lot, the science is really strong, but ...but we need decision theory, although again we come to the same thing - that our prediction it will always remain a prediction
It doesn't work!
There will always be an error in such a trading system, and the value of correct predictions will be determined not by the current deviation of the trained model and the current price (although you generally mean ZigZag), but by the number of series of continuous losses and profits
Let me explain for the last time.
There is an indicator (it can be anything) (the man wanted ZZ)
The indicator has parameters that are not constant, and everyone trades constant parameters
1) We take and find out which parameters on the history are adequate at each moment of time, get a series of values
2) we learn to forecast this series
3) at the current moment we substitute the forecasted parameter in the indicator and become adequate to the market
4) trace the error between the forecasted parameter and the real one; if the error differs, then your "the moment, when the system stopped working".
not a word about the price here!!!!!
Do you understand the importance of achieving or predicting a series of trades? - at any moment you can open a sell order and a buy order - and it will be a correct prediction! but not every order can be closed with a profit - there is a reason
Are you drunk? ) According to your statement, the market goes up and down at any time?)
Are you drunk? )) According to you, the market goes up and down at any time?
i'm not interested in discussing with the schoolboy how he made 5 bucks three times over, i'm not interested in it
ignore my posts until you learn how to use the strategy tester, I am not interested in discussing with schoolboys how he tripled his 5 bucks
is the earth even round? ))
I'll explain for the last time.
There is an indicator (this can be anything) (the man wanted ZZ)
The indicator has parameters that are not constant, while everyone trades constant parameters
1) We take and find out which parameters on the history are adequate at each moment of time, and receive a series of values
2) We learn to forecast this series
)))), at first it seemed to be worth it, but in fact it is an automation of a strategy tester. Somewhere on the forum about 7-9 years ago there was an article on this subject. Such a system would be terribly energy-consuming.
And the result will still be probabilistic :)