Machine learning in trading: theory, models, practice and algo-trading - page 3037
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
A classification error is not an indicator. The indicator is the balance and the balance line. Years 5 and more.
I showed you a balance sheet with a classification error of 8.3% on the OOS. https://www.mql5.com/ru/forum/86386/page3008#comment_46150275
Profitable, but still threw such a model in the basket.
Show your balance line with 20% on OOS. It will be an example to strive for.
I don't understand your pictures. I don't understand what it's about, what classification has to do with it, when the columns are balances.
1) No robot, as we had to reject the teacher. and there were technical problems with the counsellor. Now all technical problems have been overcome.
2) I consider your idea with balance to be unworkable, although I liked it initially. Balance can NOT be a teacher, because it does NOT exist. You have to design a teacher more carefully than I did before.
1) well then what is there to talk about?
2) And I never said that the balance idea is working, in fact I said the opposite, FF has infinite possibilities for variations.
Have you forgotten how it was? I'll remind you.
The man asked how to train the network for balance, I gave you a hint, you got interested, asked for an example, I gave it to you. I also wrote that you should not train for balance.
And all the rest is purely your personal inventions, inventions that you for some reason asociate with me.
1) well, then what is there to talk about?
2) And I never said that the balance idea is working, in fact I said the opposite, FF has limitless possibilities for variants.
Have you forgotten how it was? I'll remind you.
The man asked how to train the network for balance, I gave you a hint, you got interested, asked for an example, I gave it to you. I also wrote that you should not train for balance.
And all the rest is purely your personal inventions, inventions that you for some reason asociate with me.
The idea of balance is new to me and came from you, which I emphasise. But you're somehow reacting nervously . . . .
Closed the topic with the balance.
rule sustainability is as unworkable on OOS as the balance curve is fiat
I've done all this before, in different forms, many times....
But I still think that everyone should know how to write FF and use AO...
I'm not a fan of it either, but Alexei says it works.
The question is why he doesn't use what he got ) apparently the result is not very satisfying.
my logic for the last month and a half (it has been working for more than a year with the same success).
I consider it more successful (working with model errors) than fiddling with rules.
because fiddling with rules is the same as genetics, selecting parameters from successful passes.
H.Y I won't PR the last article anymore, because I'm bored :) If at least someone would figure it out and suggest what else could be improved, it would be progress.
I'm not really a fan either, but Alexei says it works.
the question is why he doesn't use it )
So he said many times that the rules are dying and resurrecting random is not working, I confirm this ....
Also we can conclude that if already selected rules do not work, then just AMO is not working at all ...
From this conclusion that the problem is not in the AMO, but in the data and / or targets
nice picture )
there's a little book on all sorts of rules.
https://christophm.github.io/interpretable-ml-book/
I don't understand your pictures. I don't understand what it's about, what classification has to do with it, when the columns are balances.
the balance line on the chart, in my opinion, although profitable, but not suitable for work.
The Err column shows the classification error. For the profitable variant it is equal to 8.3%, for the variant that worked at 0 the error = 9.1%.
So can you show me your balance chart on the OOS with 20% error?
What is not clear there? Let me explain:
the balance line on the chart, in my opinion, although profitable, but not suitable for work.
Err column - shows the classification error. For the profitable variant it is equal to 8.3%, for the variant that worked at 0 the error = 9.1%.
So can you show me your balance graph on the OOS with 20% error?
What does balance have to do with classification error?
What does balance have to do with classification error?
Exactly, nothing. It's unclear why you keep touting 20% as an achievement...
Neither 20%, at 8% at 50% mean anything. The numbers are about nothing.
The balance is interesting. No graph?
I'm not really a fan either, but Alexei says it works.
the question is why he doesn't use it ) apparently he is not very satisfied with the result.
It turns out to select rules, but we should understand that some of them do not work steadily from year to year, some of them stop working at all, and the other part continues to work steadily.
Of course, we are interested in those that continue to work - what distinguishes them from the others is the mystery that will significantly improve any TS.
That's exactly what I'm trying to increase potentially the number of good rules, by selecting a limited number of sections of predictors for them. In order to do this, we need to identify areas of "stable" performance of each predictor that will be used to create rules. This is the task I am currently interested in.
I have not experimented with other targets, as I am looking for a less costly way than I have for mining these rules.
Did I understand correctly that you decided not to do a colab to compare the two methods?