![MQL5 - Language of trade strategies built-in the MetaTrader 5 client terminal](https://c.mql5.com/i/registerlandings/logo-2.png)
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
thecore,
all this mathematics is very well done on NS
For example, in a probabilistic network, by changing (optimising) the sigma generalisation parameter, we get the same "clouds".
The same way you can input time (night/day) and any other data you think is important.
For example, for simple MA of period N it is also bar N (which will be dropped out of calculation of MA on the next bar).
All in all, very handy.
Sorry for the delay, but all sorts of important things prevent me from participating promptly.
to thecore
Классическое предсказание MA (в моем понимании) - это примерно так:
(1) use MA[2] and MA[1] to calculate the difference of MA[2]-MA[1] or the angle.
(2) go further to the left and look for the same angle on the history
(3) from this found point take as many bars FORWARD as we want
(4) write the average of all found values into the array
(5) go as many BACK bars through the history as we want, but it is desirable to change trends several times during this time
(6) as a result we obtain an array of averaged prediction points
Unfortunately, I haven't understood it very well already from the highlighted point, and my curiosity doesn't allow me to "score" on this method. So, calculated MA[2]-MA[1] (I understand that MA[1] is "current bar-1"), went "into history" and found the same difference (it must be exactly the same or there are some criteria). Up to this point - I seem to have got it right. Point two "settles" us on some bar in the history. Then "go forward as many bars as we want" (and if we don't want it :o) - just kidding, there must be a criterion) - we sort of look what happened on the average after the similar MA[2]-MA[1]. The fifth point isn't very clear, what are we looking for in it and why are we going into history, we've just come from there, haven't we? Or are we iteratively looking for behavior of MA after the rest of the same MA[2]-MA[1]? Is the presence of trends determined simply by the sign of the difference (sign of the formed angle)?
To sum it up - have I understood conceptually correctly that all similar differences MA[2]-MA[1] are searched through the whole history and estimated "what happened" after each event?
to anubis
Going a bit off-topic: Does anybody know where I can find forecasting algorithms, e.g. ARIMA, that would be suitable for implementation in MQL ? ps without diff. levels of course -)
I've seen a lot of similar themes on the forum, try to use search. But there are other ways: you can read books or use MathLab and look there (its m-code seems to be open). But mine, IMHO, is very simple - there's not much sense to implement ARIMA for several reasons: it's a very messy implementation algorithm, and besides you can replace it with AR model of higher order. There is a theorem that uniquely equates the orders of these models, roughly speaking, ARIMA(m)==AR(n), where m and n are orders of models.
to Aleku
In my opinion, one should not look for general approaches here - optimization theory, Markov chains are
and we can bury ourselves in it for years - but we have to find the basis for the choice of the asset and line up priorities according to this criteria.
and prioritize accordingly.
...
I think, in the near future, I'll lay out a more detailed model, if there's anything to lay out (now I'm stuck with some problems). Unfortunately time is short as always, and the subject, you're a hundred percent right there, is not as straightforward as it seems at first glance.
I am not satisfied with many things in existing money management models (perhaps it should be called that for clarity), such as lot size management. Very often they are taken from analysis of trades and as a result - almost always loss-making trades of the system are taken with maximum lot.
But never mind, let's deal with it. :о)
TASK
Colleagues, before "our everything" is ruined by the crisis - help me solve a very simple problem. Suppose the Expert Advisor has just started working (the very first initialization). One, some account of some brokerage company, some amount of traded symbols, for better accuracy let it be N.)
Of course all trading environment is known. There is an initial limited deposit, we do not take into consideration deposit replenishments. One segment of the zigzag is tentatively one trade.
The Expert Advisor requests a forecast for each symbol and obtains the forecasted zigzag for some number of segments in advance, let us assume M. We obtain a total of NxM forecast segments. For each forecast segment its geometry and risk estimation is known. For simplicity we will consider only first segments of each instrument. We get the following picture (conditionally of course :o):
Taking into account deposit limitations, segment (trade) existence time, risks, trading environment we need to find such an optimal number of lots for each trade (each segment) that will result in maximum total profit of all trades.
Why not use a genetic algorithm to find the best solution?
В итоге - правильно ли я концептуально понял, что ищутся все аналогичные разности MA[2]-MA[1] по всей истории и оценивается «что было» после каждого события?
Yes, you got that right. You are looking for MA[i+1]-MA[i] differences in the history similar to MA[2]-MA[1] or even MA[1]-MA[0].
As I said before, the use of MA is not obligatory. What is important is the principle of prediction.
TASK
Taking into account deposit limits, segment (trade) existence time, risks, trading environment, you need to find the optimal number of lots for each trade (each segment), which will result in the maximum total profit for all trades.
Hello, Sergey.
You must be joking - if I remember correctly, Markovits received the Nobel Prize for this solution (the problem of the optimal portfolio)! Would you like to find him on our forum?
You must be joking - if my memory serves me right, Markowitz received the Nobel Prize in his time for solving this problem (the optimal portfolio problem)! Would you like to find him on our forum?
Not him, but her. :-)
Not his, but hers. :-)
Nah, not hers. ^_^