Machine learning in trading: theory, models, practice and algo-trading - page 2770
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
Formalisation has never been easy.)
and then there are interest rates and related problems, in particular overnight rates. Following the Fed rate hike, Europe will have to raise rates too, but there is too much foreign (for Europe) money "overnighting" in Europe. Which in 8 hours will sleep overnight in the next place. And they will all get a piece of the pie (and the discount rate is also monetisation of GDP, i.e. a significant part of the stunted GDP will fly away for nothing).
all of this should be invested in the MO, NN somehow at once, and not expect it to identify, categorise and train itself from the bar/potential history
A new candle has come, we adjust (move the window) and again forecast the next candle, etc.
.
trading system - 13 years - during which the author proved to himself:
non-stationarity of the price series....
deterministic component and noise in theprice tried to put it into the regression equation (in EViews).
3 years later I got a conclusion (allegedly invented a bicycle):
We have a primitive regression model. It is shown that within-sample it has a profit factor much greater than 10. Out-of-sample, it is a little more than 1 and that is questionable. This model is "correctly" constructed. there are a number of indications of "correctness".
There is an idea in the model used: we isolate the deterministic component and add noise to it
after 5 years I decided to apply a new hi-tech tool(random forests) to realise the same pseudo-prediction, which is called a model... by the way forests are very resource-intensive algorithms (in terms of RAM)
ZigZag with the parameter "distance between reversals" equal to 0.0035 dollars was used to construct the target variable.
after 8 years changed the language - "Created a topic on formulation of technical specifications for MT4/5 connection with R".
although the Library of interaction between R and MetaTrader 4 terminal is available in CodeBase: mt4R for new MQL4 . https://www.mql5.com/en/code/11112.
BUT except for classDist and entropy he couldn't connect anything from new libraries...
and still shouting about his tools, his pseudo-"models" and telling everyone how to live.... the other one asks everyone to shut up, the third one promotes the first one, drawing horror stories about the 2nd one, dreaming that the 1st one will look even better against the background of the 2nd one to unite with it into one community.... and the rest don't give a damn...
13 years already (only the faces change) - and there is no working bicycle, only fairy tales about it - that"The article will be useful for both beginners and experienced traders "... and all on slogans, all on slogans... and not on logic and evidence (not their lack of evidence).
and you could have just thought first (applying Knowledge+Experience) so that you don't have to go through tools (and signs) for something that won't work apriori... and not writing rubbish for 13 years - giving it the pseudo-title of "articles"... even copy-writing contains more truth than such articles and such pseudo0exchange of pseudo-experience of pseudo- and non-colleagues in pseudo-study of anything ... and in essence
===
brand name? The engine of progress? -- all they could do was to disprove the hypothesis that their trading toy is more important than interest rates on the interbank market
When we try to create trading systems for financial markets, we constantly have to move through the territory of the unknown, step by step, gaining metre by metre some knowledge that is still insufficient.
This is a common situation and there have always been people who were able to make, or try to make another step, make a mistake, get better and go forward again.
But there are always people around, extremely intelligent, encyclopaedically educated, who throw terms right and left.
BUT
On the turn it turns out that such people are not only NOT capable of any synthesis on the basis of their knowledge, but on the turn do not understand the meaning of the words they say.
When I studied, in my group there were 6 such people out of 25. All of them had red diplomas, but they were completely impotent in the sense of realising any, even primitive, design idea.
Then I saw such people in the research institutes where I worked.
They are smart fools.
Despite the degradation of the last 30 years, these people have survived and continue to analyse others instead of creating their own.
These smart fools are more hopeless than non-scientists, it is impossible to teach them, they already know everything and with a scholarly look they drive and drive nonsense.
step by step, conquering metre after metre of knowledge... and with a scholarly look, they talk and talk rubbish.
and that's the problem, every word of it. just like a coin tossed - it falls like that -- everybody is a fool, and you don't know what to do with metres((...). the quality of your conclusions is clear
Curious note in favour of Sanych's retraining and stuttering window
https://medium.com/towards-data-science/ai-in-industry-why-you-should-synchronize-features-in-time-series-f8a2e831f87
By the way, about difficulties with windows - you can take equi-volume windows. A window with a total tick volume of 1000 bars, for example. It is a little bit closer to the physics of the process than just a window of fixed time (number of bars).
The same with tick chart is more complicated - ticks that move the quote more than tickSize or simultaneously change both bid and ask will have to correct tick_volume.
only it will not be easy to fit it into NN.