Machine learning in trading: theory, models, practice and algo-trading - page 1213
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
For about a year and a half, Alyosha gave me advice not to use ZZ on either input or output, as they peek into both past and future. Alyosha did not explain the details - here is my interpretation, how everything happens:
When using ZZ as a target, it is the past, by which this ZZ is built, that is built by previous bars (from the zero point), which are fed to the input. In other words, knowing those bars and ZZ (partially based on them) you will get a peek into the future at the input.
When I started using price increments as an output, everything became less rosy than with ZZ.
If this is the case, I don't know what to do with it. It's unreal for a beginner in MO to read 1200 pages. And once you read it, you'll miss it if there are no explanations.
That's great, if the past matters - that's what we need! ZZ has a property to describe past events - that's what I use it for, and the target is plotted as a given length of the segment from the beginning of the ZZ ( Donchian channel crossing point) - earlier attempts to guess the trend look like knife fishing, but I'll try them too.
So that's great, if the past matters - that's what we need! ZZ has a property to describe past events - that's what I use it for, and the target is plotted as a given length of the segment from the beginning of plotting ZZ (Donchian channel crossing point) - earlier attempts to guess the trend look like knife fishing, but I'll try them too.
By feeding ZZ to the output you indirectly give the model a notion of the future. The model will be retrained. Are Train and Test very different?
Of course they are very different - it is not the first week I've been trying to find a method to identify the pattern of behavior of the model on the test and independent sampling.
I do not feed the built ZZ on history, the readings are taken only at the moment of formation of predictors on a particular bar, i.e. the information is taken in the Expert Advisor, so I do not understand, what peeping are we talking about here?
Of course they are very different - it is not the first week I've been trying to find a method to identify the pattern of behavior of the model on the test and independent sampling.
I'm not feeding the built ZZ to the history, the readings are taken only at the moment of formation of the predictors on a particular bar, i.e. the information is taken in the Expert Advisor, so I do not understand what peeping we are talking about here.
The practice is a criterion of truth. Instead of ZZ supply the increment and the error will equalize on Train and Test (unfortunately - most probably by 50%). At least take off the rose-colored glasses and start looking for more real things.
Practice is the criterion of truth. Apply increment instead of ZZ and the error in Train and Test will equalize (unfortunately, most likely by 50%). At least take off your rose-colored glasses and start looking for more real things.
You do not seem to justify your statement, I have many predictors, there are also with the metric of price increment - there are different.
Justify, please, about ZZ, maybe I really do not understand something ... Preferably with pictures :)
You do not seem to justify your statement, I have many predictors, there are also with the metric of price increment - there are different.
Justify, please, about ZZ, maybe I really do not understand something ... Preferably with pictures :)
I myself was prompted by Alyosha without explanation. My interpretation in the first post. Maybe I misunderstood. But the practice has proved to me that this is so.
I myself was prompted by Alyosha without explanation. My interpretation is in the first post. Maybe I misunderstood. But practice has proved to me that it is.
Thanks for your willingness to help!
The regular ZZ has a disadvantage associated with redrawing past bars, maybe that's what you were talking about - I apply ZZ devoid of this disadvantage.
ZZ is just a way to describe the market movement on history and a point to make a decision about entering the market. With the first one it is not clear what may be wrong - there is no peeping as such there by definition, because the bars are taken with an offset and on the current bar, and as for the point of making a decision about entering the market - the question is more debatable - I consider the change of the vector of the current segment ZZ as a probability of changing the trend - so I make a decision about entering the market immediately after the occurrence of this event. The entrance itself may be made a bit later.
ah... zigzag... That makes sense then... why there are the same errors... I missed it.
Who even came up with this zigzag crap, I wonder... where the roots come from
I think this is probably the first thing that comes to mind - what to teach the model ... and here you can see the vertices and troughs, so the zigzag is the perfect solution :)Ksan Ksanych and his grandson Kesha came up with it. Who else?!
Unfortunately I had to read all 1200 pages (at that time they were 800, and then as they come in once a week), when I did not know that there was nothing much to read, I saw a little showdown about ZZ, increments, returns, etc. I think it's facepalm, shame. In my humble opinion, some gloomy characters (Alex, toxic, combinator, etc.) are just whining and demotivate Forex community, because to have doubts, to lose faith and spirit - is like death to a trader, and these whiners and direct us to this. Such provocateurs should simply be ignored.
It does not matter what is the entry and exit, if there is a dependence, the MO will find it, and I think no one would argue ZZ, it would not be bad to know where it is directed, it is a beautiful trend indicator, I see no problem, of course it looks into the future, but how else? THE EXIT IS SUPPOSED TO BE LOOKING TO THE FUTURE. The entrance should not look ahead, the exit should. A lot of articles written, both local and Western, ZZ, as an exit, and you know, if you think that everybody is wrong and you're right, then...
ah... zigzag... That makes sense then... why there are the same errors... I missed it.
Who even came up with this zigzag crap, I wonder... where its roots come from
Although it's probably the first thing that comes to mind - what to teach the model ... and here you can see the vertices and troughs, so the zigzag is the perfect solution :)
Well, well, well, what does this have to do with the trade model and the model to find good models? Or have you decided, that I've placed ZZ there by some miracle - I can't even imagine, how it could be done...
Practice is the criterion of truth. Instead of ZZ supply the increment, the error in Train and Test will equalize (unfortunately - most likely close to 50%). At least take off the rose-colored glasses and start looking for more real things.
If you do not know what to do with the market, then you will not be able to do anything with the market.
But even a 95% accurate prediction on the OOS will not save you from the sinking of the real, the market is the market, it changes frequently.