Machine learning in trading: theory, models, practice and algo-trading - page 1476
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
I don't have to say anything at all, would that make you feel better?
Okay, let's say thinned ticks can be predicted, but what's the point of predicting 1-2 ticks if it's the size of the commission.
Why not try to thin larger timeframes, for example with zigzags or something similar, leaving only price function extrema, or employ spectrum analysis for the same purpose?
Okay, let's say thinned ticks can be predicted, but what is the point of predicting 1-2 ticks if it is the size of the commission.
Why not try to thin out large timeframes, for example with a zigzag or something similar, leaving only extremes of the price function, or connect spectral analysis for the same purpose?
That's what I'm saying, that's exactly what Doc ran into - on thinned series had high predictive ability, but everything was eaten up by the spread and commission. So, he started thinning more and more... In doing so, I'm afraid he lost the nonlinearity of time, went to pure SB and got frustrated... Alas, this is also possible...
So come onAlexander_K
Or argue my wrongdoing with proofs (a screenshot of someone else's trade without nicknames is not proof of your rightness)
This is a serious accusation, passing off someone else's code as your own and showing someone else's equity is probably the most vile and condemnable thing you can do algotrader, for that eternal sleeves.
This is a serious accusation, passing off someone else's code as your own and showing someone else's equity is probably the most vile and condemnable thing you can do to an algotrader, for that eternal sleeve.
:)) Don't worry, master of cool posts in style and content - the signal is mine, I just hid it so as not to excite the minds of those who suffered.
Look, it's not that complicated...
You take the price
apply a zigzag or something similar, I use a different thing (Perceptually Important Points)
then you leave only extrema, you clean the intermediate values
you get 17 points out of the original hundred. Now you differentiate.
Now we get your favorite "stationary" series with normal ranges of tens or hundreds of points.
My network works much better with such data than with any others I've tried, but I've tried many things.
And then you can cut there or do whatever you want with it.
Look, it's not that complicated...
you take the price
you put a zigzag or something similar, I use another thing (Perceptually Important Points)
then you leave only the extrema, you clean out the intermediate values
you get 17 points out of the original hundred. Now you differentiate.
Now we get your favorite "stationary" series with normal ranges of tens or hundreds of points.
My network works much better with this data than with any other I've tried, and I've tried many things.
Well, great. Do I have a problem with that approach? Open a signal - let the suffering people earn, too.
Well, fine. Do I have a problem with that approach?! Open up a signal, let the hungry make money, too.
I'm showing you how you can increase the point spread in your ranks, I don't need it... I work with levels and do not recognize anything else (yet), and consequently I do not recognize your approaches because they (conversions) delete all information about levels
But, in fact, almost no one in this thread deals with preprocessing of input data. And yet Koldun said 1000 pages ago that this is the most important and this stage is the most important secret of all the masters of the MO. And Koldun should be able to listen.
The irony is that no matter what the Master says, he is on the contrary trying to mislead, or just to demonstrate once again the Purity of his unsurpassed Genius.
Also, the irony is that no conversions can teach better than the original quotes without any conversions. I have given examples. The problem arises from the need to scale the data, but any scaling kills the information. It is a vicious circle and it is not so easy to get out of it.
The irony is that no matter what the Teacher says, he is trying to mislead, or simply to demonstrate the purity of his unsurpassed genius.
Also, the irony is that no conversions can teach better than the original quotes without any conversions. I have given examples. The problem arises from the need to scale the data, but any scaling kills the information. The result is a vicious circle, the ring of which is not easy to get out of.
The Sorcerer is probably fooling around here and there, so he is a sorcerer. But he helped me figure some things out. In general, I'm skipping your shit - because just the presence of both of you on the forum gave me strength and I'm grateful to you for it.
As for data thinning/pre-processing - I stand by my opinion: it is necessary. But, let it be a debatable question.
The Sorcerer is probably a bit of a trickster, that's why he's a Sorcerer. But he did help me figure some things out. In general, I'm skipping your shit - because just having both of you on the forum gave me strength and I'm grateful to you for that.
As for data thinning/pre-processing - I stand by my opinion: it is necessary. But, let it be a matter of debate.
It's not a quarrel, just a quirk )
Pre-processing is necessary, but different types of differentiation somehow severely degrade all indicators, fractional differentiation included. Maybe it's necessary to make exactly thinning deals for NS and pick good sets