Methods of carrying out a rolling forward - page 11
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
Any aeroplane is based on a working simple model. Trying to build an aeroplane out of rubbish (or any number of non-flying bikes) with any number of degrees of freedom will not succeed.
Systems evolve from simple to complex in an evolutionary way. The synergistic path is of course possible, theoretically, but requires enormous resources.
Any aeroplane is based on a working simple model. Trying to build an aeroplane out of rubbish (or any number of non-flying bikes) with any number of degrees of freedom will not succeed.
Systems evolve from simple to complex in an evolutionary way. The synergistic path is of course possible, theoretically, but requires enormous resources.
I mean that it is impossible to optimize separately a couple or three variables in isolation from the system as a whole, so volking must optimize all together and there is no way to save money on it.
Except that evolution does not contradict synergy at all and in real life leads to it. You can't do synergy without selection and in that sense
you need exactly the same resources for both. Synergy in this situation is not the correct term at all, I'm actually talking about the tunability of the system.
But this is an OFTOP, I don't want to discuss it here.
And in general, when volking, we need to understand WHY we are doing it, do we want to-
1. find out if the Expert Advisor/Code Ais better than B (optimization by minutes,the absolute value of forwards is not important, the relative one is)
2. to find outthe optimum step of the history size for reoptimization or backwards- forwards ratio, (optimization by minutes, the absolute value of forwards is not important, it is relative)
3. find out whichoptimization criteria are better (optimization by minutes, the absolute value of forwards is not important, the relative value is important)
4. or prepare an Expert Advisorfor trading on Demo/Real. (optimization by ticks, you do not have to go through the whole history, you can use only a few steps, but the most recent ones)
What I mean is that you can't optimise a couple of variables separately from the system as a whole, so the volking has to optimise everything together and you can't save money on that.
Except that evolution does not contradict synergy at all and in real life leads to it. You can't do synergy without selection and in that sense
you need exactly the same resources for both. Synergy in this situation is not the correct term at all, I'm actually talking about the tunability of the system.
But it's OFTOP, I'd rather not discuss it here.
The main goal of a football forward is to score goals, preferably in a championship.
We take a 6-8 year old boy, send him to the championship and he can't do anything there. Is the boy unpromising?
It is not about Volkings, it is about optimisation as such, before the system works it is necessary to localise the ideas by which the system will work.
We need to select boys for football school based on circumstantial traits, speed of jumping, vision of the pitch. And then train them to hit the ball, and then hit the ball into the goal.
I'm talking about custom FF selection and fields of possible parameters (since you have so many parameters).
The main aim of a football forward is to score goals, preferably in a championship.
We take a 6-8 year old boy, send him to the championship and he can't do anything there. Is the boy unpromising?
It is not about Volkings, it is about optimisation as such, before the system works it is necessary to localise the ideas by which the system will work.
We need to select boys for football school based on circumstantial traits, speed of jumping, vision of the pitch. And then train them to hit the ball, and then hit the ball into the goal.
I'm talking about custom FF selection and fields of possible parameters (since you have so many parameters).
Nice analogy with footballon. You might have a talented boy, but you can never understand how he plays as a team without using him in collective play situations with 10 other players, and of varying skill levels. You are the coach and your ultimate goal is to win the team as a whole. If the stars were winning on their own, football would be a circus. Forward is about the number of goals scored, not the success of nurturing talent.
While we often know little - how the game itself is actually played - it's usually a black box. You put the team on the field with a wolfing forward and go to bed - and then they just tell you what score they played with and that's all you know about the course of the game. So you change your boy to another or give him the task of defending instead of forward and go back to sleep and compare the new score with the old one
WF is definitely the final judge of the robustness of your strategy. But you can do a simpler and not much worse on the knee - one big forward, for 7 years (well, at least for 5).
If the strategy learns noise, there will be a flop, full stop. If there are profitable parameter options on the forward, and even more so with normal performance, then you will see that.
Also one big forward will show a plum if your strategy will have too many degrees of freedom and as a consequence a fit.
It's a good analogy with football. You may have a talented boy, but you can never understand how he plays in a team without using him in collective play situations with 10 other players, all of them at different levels. You're the coach, and your ultimate goal is to win the team as a whole.
Oh, here come the cool metaphors )) Yeah, but if none of the boys hit the ball, their crowd is no better ). So start with one boy, let him drive the ball and shoot an empty net, then add a partner, train crossing, then build a team game forming models of attack, defense, etc. The process is evolutionary, don't just let the team out, make them run for the ball and go to sleep, then analyse which kid's foot put the ball in the opponent's goal and whose back prevented them from scoring in yours.
Volkings are an imitation of reality, in reality you have a professional team against you. Training an individual player is not volkking, it is optimisation, and the indicator does not earn money on its own.
During volking you can release either one genius - one man is no match for all - or one hundred cripples, getting in each other's way - the opponent will still beat you in both cases.
In short, try it your way - but I got my bearings during the hard fights with the wolves. He made me do it, though I tried to outwit him.)
Also one big forward will show a plum if your strategy has too many degrees of freedom and as a consequence a fit.
There is no connection between degrees of freedom and fitting. Fitting is the lack of forward checks. It doesn't matter how many optimization parameters you have - if you say that your EA works as shown on the optimization chart without showing any forward looking check, that's the fit.
There is a connection between the degrees of freedom and the risk that these freedoms are unreasonable, i.e. they do not affect each other and do not build feedback systems.
I.e. degrees of freedom may be interdependent or may just end up with empirically found constants and then, indeed, while we artificially adjust these constants, everything is beautiful on the back side, but once entering an unfamiliar forward environment, on the contrary, they become stubborn spoilers.
But if we consider variables, as variants of learning and reflexivity of the system, then the more of them on every occasion, the better. The main thing is that they don't interfere with each other.
There is no connection between degrees of freedom and fitting. Fitting is the absence of forward checks. It doesn't matter how many optimization parameters you have - if you say that your EA works as shown in the optimization graph without showing any forward checks, that's the fit.
There is a connection between the degrees of freedom and the risk that these freedoms are unreasonable, i.e. they do not affect each other and do not build feedback systems.
I.e. degrees of freedom may be interdependent or may just end up with empirically found constants and then, indeed, while we artificially adjust these constants, everything is beautiful on the back side, but once entering an unfamiliar forward environment, on the contrary, they become stubborn spoilers.
But if we consider variables, as variants of learning and reflexivity of the system, then the more of them on every occasion, the better. The main thing is that they do not interfere with each other.
We can discuss this topic. If you have 10 000 transactions (points in multidimensional space that we limit ourselves with selected dimensions - variables), it's enough to introduce 10 000 variables to fit all your points on the hyper-plane. This is already a barefaced fit. So, what I'm saying is that
There is no relationship between degrees of freedom and fitting.
That's not true in principle, but there are times when it's not critical.
There is a connection between the degrees of freedom and the risk that these freedoms are unreasonable.
validity can be detected. If changing some variable with the others fixed does not change the outcome much, significance is low. However, the interaction with other variables is lost. In fact, the system is an interaction of several variables.