To follow up - page 28
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
Hmm, I thought I had a rough idea of what you were talking about, but with this post you've completely shattered that misconception :). But I'd like to get to the bottom of it. How about we start in order, in small steps? Using one parameter as an example.
So, here is the price series, you go along it and calculate the parameter at every bar. Or on every tick?
Then you build a phase space from these values. You said that the phase trajectories should be continuous. So you take ALL values? From every bar (from every tick) ?
If not, how do you determine at which points in time to take values for further analysis ? And in what sense then did you talk about continuity?
Let's take it in steps. The parameter, imho, must be local, otherwise how can it describe the market condition. Consequently, it must be calculated at each datum (it doesn't matter if it is a bar or a tick).
The phase space is not built from the values. The phase space is the entire range of values of a parameter. If there is only one parameter, the FP is a line (or its segment). If there are two, it is a plane (or its region).
The phase trajectory is continuous - this means that it has no discontinuities of either the second or the first kind. For continuous functions, continuity is uniquely defined. In our case, the process is discretized. Therefore uniqueness is not achieved here and continuity has a very approximate meaning. Nevertheless, we both understand that it means that the changes of a parameter on each datum are substantially limited (you can do it in a statistical sense). So the question of whether I take ALL values sounds somehow unclear to me. There are exactly as many values (i.e. experimental points) as I take for analysis. But the FP points are slightly more. :-)
So, if we draw a long enough trajectory, we don't yet have anything for context extraction from it. We have to specify entry points and exit points (or entry points to the opposite position) on the trajectory. This is what a completely external algorithm does. In my version of an ideal (i.e. focused on maximal profit) system, these are the tops of GZ. In your version, these are the entries and exits, which are set by your strategy of opening and closing positions. And you use it only because you subjectively estimated it as leading to a profit with high probability.
On the one hand, we get the market parameterization that leads to such or other trajectories, and on the other hand, we get the entry-exit algorithm. If from their connection the input points form a cluster in one place and the output points in another, this is the desired result. In your case it is the end result. In mine, we should check whether the parameter values alone are now enough to open and close trades. If yes, then it is also the end. If it isn't and the mere finding of a trajectory in the long cluster is not enough to open the corresponding position (in your language - in this cluster there is still too high a percentage of "bad" points), then we should look for an additional entry filter. Or a more complex algorithm of inputs. Or another parametrization.
I hope you have understood, that concept of trajectory continuity and input-output timing are two "big differences". :-))
учитывает еще и гладкость эквити - например ПФ. Тогда .....
PF is an interesting thing...... let's take a period of 12 months (a year) .... in the constraints we put the deposit 1000000 and the percentage 0.08..... we get 3-5 thousand parameter variations...... we test (no matter how) and no matter what parameter ..... we get the same thing - 800, but for 2-3-4 years..... and what to do with it? ..... with PF.....
the only use I have found for myself is in tuning the MM
I'm a beginner, man :(((Надеюсь ты понял, что понятие непрерывности траетории и моменты времени для входов-выходов - это "две большие разницы". :-))
The problem wasn't that I didn't understand it, the problem was that I didn't understand why you needed continuity. And I still don't, because it's not used anywhere yet. What do you care if your cluster consists of discrete points or points belonging to some continuous trajectory.
However, judging from what is written it has nothing to do with understanding your approach, so far it is exactly as I imagined.
OK, let's move on. So you've overcome the parameterisation problem and have the coveted clusters for perfect entries and exits. How are you going to checkwhether"the parameter values alone are now enough to open and close trades"?
to Night....
for any market there is one and only one decision: buy, sell, refrain ..... how this decision is made is a tenth matter.... but the way to figure it out is the first thing..... it can be done by experiments or market research, but it can also be done by optimizing an expert based on this idea..... you may say that it is not "context" at all - whatever the context, when an idea is profitable.....
I'm flubbing in honour of Christmas..... and Happy Holidays to you :)
The problem wasn't that I didn't understand it, the problem was that I didn't understand why you needed continuity. And I still don't, because it's not used anywhere yet. What do you care if your cluster consists of discrete points or points belonging to some continuous trajectory.
However, judging from what you've written it has nothing to do with understanding your approach, so far it's exactly as I imagined.
If the value of a parameter can jump at any moment from a long cluster to a short cluster in one count, who needs such a parameter? This greatly complicates working with such parameters and makes the task of collecting statistics (as grandpa Lenin said) a very complicated one. If the parameter changes continuously, i.e. more or less smoothly, some time will pass until it leaves the long cluster and reaches the short cluster. This is exactly the time it takes for the profit to grow.
Not every decision requirement is directly reflected in the strategy. There are also implicit conditions. For example, not only do you need transportation to get to the train station, but you'd better have something to wear as well. It is not easy to do without clothes.
By the way, continuity is not the actual continuity of the trajectory. Counts are discrete, so trajectories are a discrete set of points. But I will not repeat myself, I have already written about it in a previous post. Read it and you will see. And a cluster does not consist of individual points or a piece of trajectory. A cluster is a part of an FP. I think it's obvious.
OK, let's move on. Here you've conquered the parameterization problem and got the coveted clusters for perfect inputs and outputs. How will you checkif the"parameter values alone are now enough to open and close trades"?
Если значение параметра в любой момент может прыгнуть за один отсчет из кластера лонг в кластер шорт, то кому такой параметр нужен ? Это колоссально усложняет работу с такими параметрами и делает правильную постановку задачи для сбора статистики (как говорил дедушка Ленин) архисложной. Если параметр меняется непрерывным, т.е. более-менее гладким образом, то пока он выйдет из кластера лонг и дойдет до кластера шорт пройдет некоторое время. Как раз то самое, за которое вырастает профит.
So now you start testing? OK. You've done it and it turns out (since your clusters are long) that the inputs and outputs are noticeably different from the ideal. Moreover, some inputs turn out to be missing. But that's nothing, it's worse that some of the outputs are skipped too. In other words, you miss the turnover and lose your carefully nurtured profit. But that is not the real trouble. The real trouble is that your entry and exit conditions are fulfilled in completely unplanned situations. You haven't insured yourself in any way against hitting clusters of extraneous points.
What's your action?
P.S. By the way, you wrote that you count the probability of a positive outcome. Will you do it after the test results?
All the questions you asked have their own more or less intelligible answers.
However, all these questions, on the other hand, go far beyond the concept of FP and working with context.
Is there a need to go beyond the scope of the topic ? Or is it your own methodological problems ? Maybe just curiosity ?
In short, I'm a bit confused as to what you're getting at and where you're going with it. :-)
The fact is that I started out with exactly what you call your system. And my system was precisely my answer to these and other serious problems of "your" system. So I'm really curious if you've found satisfactory answers by staying within it. Or do you just believe that someday you will solve problems by finding the "right" parametrization, by some miracle? Or have you not realized them yet and therefore do not understand the fundamental difference between the approaches?
I really don't understand the fundamental difference in approach. If you can, explain.
What has been discussed here - the concept of FP and its clustering - is not yet TC and even less so MTS. It is still only a methodological framework, in my view the right one, for market research and analysis of the results of this research. Certain aspects of this concept can be used in TC, including - directly. But to expect it to "solve all problems" would probably be a bit naive. The correct methodology is good, but the solution of fundamental questions is the result, first of all, of creativity (and not a miracle). And it is beyond all methodologies.
Therefore, staying within the framework of the mentioned concept, one will obviously not find all the answers. And what I have found is also not directly connected with FP, however it provides its parametrization and can be used exactly with the help of the described approach.
Видимо вы имели в виду что со временем деформация поверхности может привести к смещению оптимальной зоны.
Yes. And assessing the impact of the "time" parameter is probably the most creative part of the whole operation. The good thing is that the optimal surface area is potentially a more stable entity than the optimal area on a line, for example.