Is there a pattern to the chaos? Let's try to find it! Machine learning on the example of a specific sample. - page 24
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
The simpler ones - you say they don't work.
It just turns out that the nearest transition bar will be classified negatively, and the next one positively - and the predictors will not change much during this time, which complicates the training.
It doesn't work well for you either. So it makes no difference.
But I'll be looking for something else. I don't want to sit in a drawdown for 2 years.
It's not working well for you either. So it doesn't matter.
But I'll be looking for something else. I don't want to sit in a slump for two years.
I haven't written about the results of this approach yet :)
Preliminary - the averages are better.
I just see a clear problem, apart from others - change of probabilities for selected quantum segments on different samples, I want to think how to detect such quantum segments better, which should already improve learning.
There is also such a problem - what to do with indicators on the readings of which predictors are built, even the same ZZ if you take, the settings can be selected to improve the predictive potential of the sample. Is it worth it, or is it a fitting, and if it is not worth it, how to fix the value of indicator settings? For example, I use oscillators with default settings.
There is also such a problem - how to deal with indicators on the readings of which predictors are built, even the same ZZ if you take, the settings can be selected to improve the predictive potential of the sample. Is it worth it, or is it a fitting, and if it is not worth it, how to fix the value of indicator settings? For example, I use oscillators with default settings.
You can take several ZZ with different settings. And indicators too. Although you already have 5000+ predictors..... you don't need much more than that.
You can have several ZZ's with different settings. And indicators too. Although you already have 5000+ predictors..... there's no need for more.
I've now added ZZ predictors on each TF up to an hour.
I tried to adjust ZZ by the number of good quantum segments from each setting, but there are doubts how to take into account similar signals actually - you need additional checks at least for correlation. And such checks are costly, plus it is not clear how to assign uniqueness to which ZZ tuning, if correlation will be revealed.
Even got curious, how do I do the math?
a long time ago in the discussion of Dimitrievsky's publication I described it in detail.
and now I described it, but the site glitched and the answer did not go through. I guess it's not fate :-)
Now added ZZ predictors on each TF for up to an hour.
I tried to adjust ZZ by the number of good quantum segments from each setting, but there are doubts how to take into account similar signals in fact - we need additional checks at least for correlation. And such checks are costly, plus it is not clear how to assign uniqueness to which ZZ tuning, if correlation will be revealed.
some time ago in a discussion of Dimitrijevsky's publication detailed.
and now I described it, but the site glitched and the reply didn't go through. I guess it was not my fate :-)
Yes, bad luck - it's always a shame when your labours are lost.
Let the catbusters do the math. I think it's faster than calculating correlations. And most importantly, he will choose what he needs.
I showed above that it is difficult for him to choose what he needs, even though it is there and he can't see it :(
I prefer easier targets. I mark TP/SL for each bar. The model itself decides which of them can be successfully traded.
So if you have TP more than SL, and the accuracy is around 50%, then you can fix the financial result at the expense of the lot size and take the proportion only in money.
I just thought about it, looking at my own strategy (the screenshot of which I showed above) - I have a difference of 61.8 per cent (ideally - in fact less because of the delay in decision making).
So if you have TP more than SL, and the accuracy is around 50%, then it is possible to fix the financial result due to the lot size and take the proportion only in money.
I just thought about it, looking at my own strategy (the screenshot of which I showed above) - I have a difference of 61.8 per cent (ideally - in fact less because of the delay in decision making).
At TP=SL it will be about 50 per cent. At TP = 2*SL, it will be 33%, etc.
Always the average profit from 1 trade is very small. About 0.00005. But it will be spent on spread, slippage, swap, which are not taken into account in the teacher's markup (spread is taken into account, but the minimum per bar, the real one will be higher).
And this Using TP=SL=0,00400. That is, with a risk of 400 we get a profit of 5 pts, i.e. an advantage of about 1%.
I would like to take at least 10 pts from the movement of 50 pts, but there all options are plum.
But this is all with my chips and targets. Perhaps there are better options.