The perfect mechanical trading system. - page 10
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
1) An extremum and an area of sufficiently good points around it.
2) Over a period of time of the order of a day or a week this extremum does not shift sufficiently, so that the selected point remains in the area around it.
Again, this may or may not be believed. I cannot justify it. If you don't believe it, it's all nonsense and there's nothing to talk about. If you believe it, you can take steps.
Shit... I really don't like the way the MT4 optimiser works. It seems to show only profitable passes in the report. And it remains unclear with what "density" this or that profitable pass is surrounded by unprofitable ones... And this is very important.
You live and learn a long time... :)
So, there is a function of many variables (for the sake of argument, this is our strategy). We need to find its extremum. But NOT ANY extremum, but an extremum smooth enough. The extremum should be smooth enough so that there are no bad points in the vicinity of the "good" point. It should resemble the bottom of a concave bowl, like a paraboloid of rotation, rather than a forest of columns sticking out of a gently descending floor. If this extremum is global, so much the better. If not, well, you can't kiss all the girls or make all the money. The main condition is not globality, but smoothness. Without smoothness it will only be interesting for history trading. I think maybe Monte Carlo is more promising than genetics here, for such an extremum should have a rather broad attractor ? One more thing. Perhaps a new such extremum should be identified in the vicinity of the old one, because everything is smooth and we assume that the market is also quite smooth on a daily scale. Who has an opinion about this ?
As for smoothness, in Metatrader when optimizing two parameters there is a property of two-dimensional optimization chart surface, green grid "x" and "y" where (as written recently in "articles" section) one can visually see smoothness (the more area and values around it are greener - the smoother parameter, and unnecessary eextremes are separate spots there". I'm not a programmer, but I think the algorithm looks like this: we take an array of "x" and "y" and write there the values of the optimization result for two parameters. For example, we will optimize by profit or profit factor (whatever is more preferable for someone). Accordingly, after optimization of parameters and recording them in the array, we will find the area that we will need. What is our parameter? We take the area of checking 9 squares of 3х3 array, (we call this area "segment") where the central square is our parameter. How to find it? We loop through the array "segments", and around it the adjacent eight values, provided that all values around the central checker are greater than zero and the sum of values in 9 cells (segment) is greater than in other segments, it is the desired segment, where the central point of the segment is the desired parameter with values of two optimisable parametres.
Thus we find a "bowl" of smooth extremums of parametric values.
Generally, we run it once a day, find smooth optimized values of parmeters and substitute them with new values for the new day. (By the way, in theory if we follow the version that the market is volatile, then if we plot values of the array of optimized parmeters, we should get a smooth sinusoidal graph).
New, an attractor (classical, not Lorentzian, etc.) is a basin of attraction. In simple terms, there is a basin. At the bottom there is a pond, on top there is a ridge. If a drop falls inside the ridge, it rolls into the pond.
The fact that an attractor is a particular point that arises when you consider bifurcations of solutions
of differential equations, I'm aware of that. I was wondering how an attractor
characterize the extremum of a multidimensional function?
Anyway, what about the attractor, now about faith?
Now the much more substantive question about statics. This is the most vulnerable part of my reasoning, and I can't hide it, I can't justify it coherently. I suggest that we simply take it on faith. This belief is indirectly confirmed by the fact that Expert Advisors optimized according to the recent history have been successfully trading for some time. To be more accurate, I suggest that they are quasi-static but not static. I.e:
1) An extremum and an area of sufficiently good points around it.
2) This extremum does not move significantly enough over time on the order of a day or a week, so that the selected point remains in the area around it.
Again, this may or may not be believed. I cannot justify it. If you don't believe it, it's all nonsense and there's nothing to talk about. If you believe it, you can take steps.
In the market one cannot trust anyone, not even oneself.
of negative and positive outcomes, primarily negative ones.
If you remember that trends on the market are changing all the time - flat changes the trend, high
volatility changes with complete calm etc. then it is difficult to imagine that for all market conditions there is only one trading strategy.
then it's hard to imagine there is ONE trading strategy for all market conditions. No amount of optimization can ensure
to make a trend strategy work profitably in a flat condition. That's why it's important to timely detect
and switch strategies accordingly, rather than trying to optimize a strategy which at the moment is not profitable in a flat market.
strategy that does not work in the current conditions.
With such a background in matters of faith, you could become a priest ... )
But seriously, here's an idea about "auto-optimization" and in connection with it, is there any way to call the strategy tester built into MT from the Expert Advisor with parameter transfer into it?
There is such a possibility.
Great, then maybe you as an mql specialist already have experience using strategy tester in such a way?
To be more specific, we call
from mql, pass optimization parameters into it, then get the results from it, analyze them and substitute into the appropriate variables.
Share your experience, if you don't mind, because this might be a dead end and it's not worth trying for a long time.
There is no need to bother with a strategy tester. All it has to do online - to optimize the parameters of the current EA - all this can be implemented inside the EA by a script, which will be called once a day or once a week, and which will loop through the different options of the parameters to perform all that the EA does, and thus use the same functions. It means that it will take very little (relatively) additional code to build all this in the Expert Advisor.
There is no need to bother with a strategy tester. All it has to do online - to optimize the parameters of the current EA - all this can be implemented inside the EA by a script, which will be called once a day or once a week, and which will loop through the different options of the parameters to perform all that the EA does, and thus use the same functions. It means that it will take very little (relatively) additional code to build all this in the Expert Advisor.
I tried to do it and encountered two difficulties:
1) The tester inside the Expert Advisor is unbelievably slow, I know it could be optimized (using your own price series, etc.), but it will still be slower.
2) For a more or less quality testing you need either your own tick emulator or
tick history saved in a file.
Plus I prefer to use ready-made blocks, modules, etc. rather than build my own bike.