Machine learning in trading: theory, models, practice and algo-trading - page 180
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
Suggest options, how can you optimize one variable so that the other comes to 0 ???? Or tends to zero.....
Generally, optimize a variable based on another variable....
F(-fabs(y(x))
right?
Please formulate your question more accurately.
The alglib library was recently added to mql, this should be solved with it.
step 1: write a function that will calculate the value of the second variable
return(MathSin(inp)); //для примера новая переменная будет вычисляться как синус из первой переменной
}
You have your first variable inp, from it will be calculated a second variable with function Qwe(). In your case, the function Qwe() needs some code, you need to look at the history of prices and calculate something there using your methodology.
Further we will face the usual optimization problem: we will have to change the variable inp in order to calculate the value of the Qwe() function. By changing inp you want to achieve a smaller result of the function. This can be done in Alglib, here is a list of tools http://www.alglib.net/optimization/ . I thinkLevenberg-Marquardt algorithm should work. I don't see any mql examples there, but the names of functions seem to be the same, you can look at examples in c++, most likely the mql code will be almost identical.
I had an idea a long time ago, maybe someone has tried it?
We are talking about "pre-life" models.
It's used in medicine.
You take a lot of information about a patient, and the model predicts how long the patient will live.
In our case.
We take a bunch of information and predict what TR/SL the price will reach.
Mihail Marchukajtes:
If we consider that the output variable is regulated by the amount of TC profit, then by changing this parameter we can at least know the quality of our input data...
The example has been laid out, and the theories should be based on it. And Andrey followed it up with the right phrase. Concerning the orientation on
The theory should be based on it, and Andrey made a correct point following it...
I had an idea a long time ago, maybe someone has tried it?
We are talking about "pre-life" models.
It's used in medicine.
You take a lot of information about a patient, and the model predicts how long the patient will live.
In our case.
We take a bunch of information and predict what TR/SL the price will reach.
The most interesting thing is that in most cases it will "get there" and there, which is often exploited by overstayers trading without stops. Until that rare moment when the deposit will not be enough to stop. In other words, it does not matter what values it will reach, but in what order these points will be reached.
This is if we sit it out without a king in the head.
And if we have a TP forecast with some decent probability, it makes sense to wait out the drawdown, and if the forecast is for a loss, it should be fixed immediately, instead of sitting and waiting for the deposit to be drained.
That's the point of the "wait-and-live" model.
No. You just find out how much will drop at another target from the same inputs. You're changing it.) Everything must be somewhat formalized.
The example has been laid out, and the theories should be built on it. And Andrew expressed the right phrase after it. Concerning the orientation on
The theory should be based on it, and Andrey made a correct point following it...
Here you're wrong, or rather right but not to the end, indeed it will give the level of entry on the signal, but the most important thing is not the number of points earned by the signal, and that this signal is not an error. This is important!!!!
Anyway, now I have adjusted the output to the level of 0.00008 pips profit. Where the number of zeros and ones is equal :-) So, I have a lot of such tricks. I have not opened all of them yet. What to do it for Tell me????
Here you're wrong, or rather right but not to the end, indeed it will give the level of entry on the signal, but the most important thing is not the number of points earned by the signal, and that this signal was not an error. This is important!!!!
Anyway, now I have adjusted the output to the level of 0.00008 pips profit. Where the number of zeros and ones is equal :-) So, I have a lot of such tricks. I have not opened all of them yet. What to do it for Tell me????
It's not "so that this signal wasn't an error", but why "this signal wasn't an error" and most importantly
that "this signal wasn't an error" in the future... etc. You have a fitting on history with an implicit attempt
volatility prediction. All imho of course...
I don't need it. And it won't work properly either.
It's not "so this signal won't be an error", but why "this signal won't be an error" and most importantly
that "this signal is not an error" in the future... etc. You have a fitting on history with an implicit attempt
volatility prediction. All imho of course...
No, get to the root of it, because what are we doing? We divide the input data, right? And so there's no bias in building a model when it knows the state "NO" better than "YES" or vice versa. NS appears to have unambiguity when building a model, without skewness. This is when the number of zeros equals the number of ones. And the most important thing is not the fact of division, but that it is constant. If you start to lose money, reverse the signal and go uphill :-)
Micha, again?))) hilarious... I don't know about you, but we share, and the regime, and wrestle shreds)))