Machine learning in trading: theory, models, practice and algo-trading - page 2374
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
Prada deal marking
Prada deal marking.
This stuff is more interesting. I just don't get it, does it only work from the command line? Has anyone looked at it?
This material is more interesting.
This is another topic, which is not limited to GANs
Pradov's Deal Marker
Unclear language and unfamiliar functions... and the author is misleading.
By fixed_time_horizon() there is this line:
idx_lower = data[data[name] < - threshold].index
he wrote above that
And the pictures below are not int (i.e. 0,1,2,3...), but 0.05, 0.01...
It became clearer with double - this is the same thing I did with TP=SL=some value of price change.
But it is unclear why I called the method and function fixed_time_horizon(). It's a fixed price change, not time.
---------
As for the method of quantized_labelling() - I don't understand anything from the code. I suppose that it is not a fixed value, like 0.05, but according to the quantile that is constantly changing with price volatility.
Unclear language and unfamiliar functions... and the author is misleading.
By fixed_time_horizon() there is this line:
idx_lower = data[data[name] < - threshold].index
he wrote above that
And the pictures below are not int (i.e. 0,1,2,3...), but 0.05, 0.01...
It became clearer with double - this is the same thing I did with TP=SL=some value of price change.
But it is not clear why I called the method and function fixed_time_horizon(); where is the fixed time? It's a fixed price change, not time.
---------
As for the method of quantized_labelling() - I don't understand anything from the code. I suppose that it is not a fixed value, e.g. 0.05, but according to the quantile that keeps changing with the volatility of the price.
I have not read the code. The main thing is the markup not based on a chart but on the increments. This leads to a number of interesting features, such as the ability to apply the markup to a squelch chart or to some specific BP components
there must be a mistake with the int, it wasn't Prado himself who wrote it, but some types
the fixed horizon refers to the selected incremental lag, I guess
I haven't read the code. The main thing is the partitioning not by chart, but by increments. This leads to a number of interesting features, such as applying partitioning to a squelched chart, or to certain BP components
there must be a mistake with the int, it wasn't Prado himself who wrote it, but some types
fixed horizon refers to the selected incremental lag, probably
Somebody there is stupid or the Prado or his types.
By the method quantized_labelling()
I see little sense in teaching it. You can learn the classification well at moments of low volatility, and it is worse at high volatility. And then a 40% error at low volatility + 51% error at high volatility will return the system profitability to 0. A lot of small gains can be compensated by a few big losses.Somebody out there is stupid or Prado or his types.
everything is zshibizzy, we should try it, but I'll do it differently
His book is a little different, I think. I'm too lazy to look.everything is cool, I should try it, but I'll do it differently
His book describes it a little differently, I think. I'm too lazy to look for it.For quantiles I don't see the point, see post above
TP=SL I tried. The result is 50% on new cross validation data.
I don't see the point in quantiles, see post above
Here are the increments, without sl and tp
I did this through clustering, marked up. Overall the curve on marked up data is not great, but more robust on new data