Machine learning in trading: theory, models, practice and algo-trading - page 1844
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
What about Maximka? Have you read anything? Any cuts?
I improved my approach a little, the result is also better... All inputs showed + :))
But there are problems...
1) there are not enough signals.
2) the model dies in time
But I think I`ve started to understand something in this damn market, so I`m sure I`ll have a breakthrough.)
Maxim LENTY :-)
Me too, to be honest. I've been contracted for an article. The topic is ours and interesting. I drew up an outline of the article and based on it I am very much interested. Already threw the first paragraph and ....... The paragraph is full, I am too lazy :-( But I will do it, because this plan written by me intrigues me very much..... :-)
Michael, please tell me what generalizations apply to price?
I would like to understand the vector of logic of reflection.
And another question.
Do you supplement the technique with exogenous characteristics?
Please give me the details for understanding.
Michael, please tell me what generalizations are applicable to price?
I would like to understand the vector of logic of reflection.
And another question.
Do you supplement the technique with exogenous characteristics?
Can you please throw these features for understanding?
The only generalization for the price is to see the trends on higher TFs from the working one. This is the essence of the NS operation. If it sees trends on М15 and H1, it works. This is, in fact, its main purpose.
For the rest, I take data on volume, delta and look for changes in these data at a certain time of the market at a certain interval (individually for each signal) I build the stochastic component, the standard deviation and the simple difference in accumulations, and that is enough. I sift out from 7500 input data up to 150 variables and use them to train 50 examples. That is the number of input variables should be about three times more than the examples themselves so that there is plenty to choose from. In this case good enough models are obtained, even if they do not work long but with high quality. This is in two words to describe....
The only generalization to the price is to see trends on higher TFs from the worker. This is the essence of NS operation. If it sees trends from H1 on M15 then it works. This is, in fact, its main purpose.
For the rest, I take data on volume, delta and look for changes in these data at a certain time of the market at a certain interval (individually for each signal) I build the stochastic component, the standard deviation and the simple difference in accumulations, and that is enough. I sift out from 7500 input data up to 150 variables and use them to train 50 examples. That is the number of input variables should be about three times more than the examples themselves so that there is plenty to choose from. In this case good enough models are obtained, even if they do not work long but with high quality. This is in two words to describe....
In general the point is clear, thank you.
You are referring to the direction of price movement.
Does not it seem to you? That the generalization of price, to bind only to a time interval, is not quite the same.
It seems to me that there should be more options.
As I understand it, you only use endogenous data and do not take exogenous data into account.
I thought you could tell me something about foundations. But as I understand it, you only build models by technique.
And another thing, I am studying now the MO from the americans. and they say that the NS models are slow in relation to regression.
What are your thoughts on this, if any.
Or it all depends on the problem, as always?
I have a suggestion to combine efforts. I've got a good entrance, but not many, so I need to trade all symbols at once, I need a robot. Since entrances are often at the very minimum, I need to enter quickly to keep advantages, I need a robot.
Karoch. there is a proposal - to create together ATC.
1) I will generate the TS in my R . TC in the form of a file with log. rules.
2) You make a tool that will open trades in MT4 or MT5 on these rules.
What do you think?
There is a suggestion to combine efforts ... As you have seen my entries are good but not many, so I need to trade all the tools at once, I need a robot, since entries are often 90% at the very minimum, I need to enter quickly to maintain the advantage, I need a robot.
Karoch. there is a proposal - to create together ATC.
1) I will generate the TS in my R . TC in the form of a file with log. rules.
2) You make a tool that will open trades in MT4 or MT5 on these rules.
What do you think?
And most importantly, all at once here, for evaluation, so to speak, by independent experts.... We may pull the branch together, but what about the corners? :-)
Max, what happened to your laziness? Tell me how you conquered it?
In general, I understand the meaning, thank you.
You are tied to the direction of price movement.
Don't you think? That the generalization of price, linking only to a time interval, is not quite right.
It seems to me that there should be more options.
As I understand it, you only use endogenous data and do not take exogenous data into account.
I thought you could tell me something about foundations. But as I understand it, you only build models by technique.
And another thing, I am studying now the MO from the americans. and they say that the NS models are slow in relation to regression.
What are your thoughts on this, if any.
Or as always it all depends on the problem?