a trading strategy based on Elliott Wave Theory - page 218
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
...
Меня интересует, увы, даже не число, а функция. Именно, зависимость вероятности успешной сделки от тейкпрофита и значение стоплосса, который при этом должен использоваться.
Но можно, в первом приближении, зафиксировать допустимый минимум вероятности успеха, при котором советник открывает позицию, и тогда просто искать уровень тейкпрофита, то есть минимальное изменение цены в нужном направлении при заданной вероятности.
The method of enumerating parameters and observing the result, otherwise there is a high probability of getting bogged down in analytics (the solution is often not guaranteed). But be prepared to be accused of "over-optimising" the system.
Yes, pure analytics is the fairytale dream of my science period. In this case - simply impossible due to lack of appropriate training. That is why there remains only the processing of results of numerical experiments on history. Actually about a correct statement of a problem in these experiments I asked.
Well, if there are no remarks to program, which I stated here: Yurixx 12.01.07 18:41, then I will try to implement it.
And the accusations of "overoptimization" are of little concern to me. There is only one objective judge, forex, and it will give the final judgement. Moreover, in the proposed formulation of the parameters is virtually absent. If the pair (BL,BR) allows to divide the whole range of values into three subfields as formulated by Neutron , then for a given value of TP we only need to construct a function of probability of a successful entering above this range of values. If it does not allow to do this, then bin it and start all over again.
Using a built-in tester, especially at this stage, does not seem to me very useful and inconvenient.
I wrote my own handler that builds an appropriate array of results on a given history.
Then, by sorting it out by one indicator, making a corresponding selection and sorting it by values of the second indicator, I get an array of data for the elementary square from the range of values. And then I can do whatever I want. I can give them +1 and -1 weights, draw a distribution of results (i.e. price changes obtained from each input), sort these results in descending order and obtain the probability for a given value of price change, etc.
At this point I have an issue with descreening the area of values. As it is reduced to a square with side [-1,1], it can be divided into cells with side length S=0.1 or S=0.01. In first case we have 400 cells, and in second case we have 40000 cells. Accuracy of surface representation seems to increase, but statistical reliability of probability values, as it is evident, falls radically. Competing processes.
How to determine the optimal cell size ?
The archive turned out to be broken (says the file is damaged), so the search continues, if you find the link before me you can also tell me what's what...
How to determine the optimal cell size ?
Maybe pages 81-84 from Bulashev can help to solve this problem?
At this point I have an issue with descreening the area of values. Since it is reduced to a square with side [-1,1], it can be divided into cells with side length S=0.1 or S=0.01. The first case gives 400 cells, the second gives 40000. Accuracy of surface representation seems to grow, but statistical reliability of probability values, as it is evident, falls radically. Competing processes.
How to determine the optimal cell size ?
The relative amplitude of fluctuations of an estimated random (pseudo-random) value dA/A is defined as:
dA/A=1/SQRT(n), where n is the number of events.
An acceptable relative magnitude of fluctuations is estimated from specific boundary conditions, but a value less than 10% can be considered satisfactory, where n>100. Thus, the size of a unit cell is chosen from the condition of sufficient number (n>100) of events in it. Of course, the same number of events will not occur on all parts of the area we are interested in. In this case, we have to make a known compromise, which will be more accuracy in the central area and not enough in its periphery.
http://www.filefactory.com/file/733226/
PS: How long it will be lying around, I don't know.
But I can't download from the file factories, it sends...
grasn if you can upload here
http://zalil.ru/
Only it will have to be split into several volumes...
grasn if you can upload it here
http://zalil.ru/
Only it will have to be split into several volumes...
There must be something wrong with the download.
I'll give it a try...
PS: Good thing I have unlimited traffic... :o)
If the factoring file sends you, it probably says something like "no slots available". In that case you just need to wait a bit and try again. Or just download it in the morning when Moscow is not yet awake - then everything is downloaded the first time. Why torture a person by splitting into several volumes?
But if you have a dynamic IP and it says that you've run out of free limit of 1Gb, then you really have only one option - "First come, first served", that is, download in the morning. In my case with a static IP I can download a guaranteed 1Gb daily without having to get up especially early.