Algorithmic ''centrifuge'' - page 12
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
If you have an algorithm for finding the "ideal" entry/exit points, then this is the indicator you are looking for, which gives the appropriate signals. Any other indicators become unnecessary in this case.
It is not difficult to attach an order sending to these signals.
But you don't have such an algorithm. Since everything in your reasoning is based on it - and there is no algorithm, there is nothing to rely on.
It isn't. The algorithm for finding points works ONLY on historical data. It shows you the best spots for trades in the PAST.
Indicators are used to predict the FUTURE.
Going back in the past, we know where the best trades were entered and exited, and retroactively select indicators that were least wrong, expecting that they will not be wrong often in the future.
Try it)))
I think I understand why there is a misunderstanding :)
Peter talks about creating an automatic search for strategies based on an indicator, without yet considering the result of this experiment. And the guys want to immediately see the meaning of this action. That is why, when asked why, Peter says it is an experiment. Right?
It is possible to build such a system of searching for strategies. I have constructed similar through saving the results of a pass into a file, with further study of this file in subsequent passes. But in 4. This is very convenient, no need to sit near a computer, charged for the whole day, and all the results are recorded.
Another thing is to look a little ahead. Is it possible to find a profitable strategy on standard indicators without wasting a lot of time? I marked a few ideal points to sell on the chart and threw some standard indicator.
What indicator values do we see in these points? There are only four ideal sell points in this timeframe. And how many similar indicator values? About 10-15. It's like that with any indicator. This is the problem.
This is not the case. The point finding algorithm works ONLY on historical data. It shows the best spots for trades in the PAST.
Indicators are used to predict the FUTURE.
Going back in time, we know where the best trades were entered and exited and retroactively select the indicators that were least wrong, expecting that they will not be wrong often in the future.
If you have such an algorithm, then this is the indicator you are looking for.
But you don't have it.
Further in a circle
The task is to automatically select indicators for a trading signal to be collected in TS, based on "ideal" entry/exit points calculated with the help of a tester ( tick history, optimizer, GA, and a special algorithm for finding such points).
ideal entry points on history - ZigZag, you have rejected
well, okay, let's say Perfect_ Entry Points_XXX
and you expect that you can find a set of indicators that can be run in the tester - but the tester is not needed here, and you can find an algorithm that will "float" values of indicator settings
imho, you won't find it, even on the not ideal, in your opinion, ZigZag - you will be able to cover only partially with this collection of indicators, but you won't be able to cover completely all peaks of ZZ - neural networks are not taken into account, they are for another purpose
I thought that at least you have the structure of the future program, I can write it myself, but the correct program layout I would be ready to use - I need it!
I think I understand why there is a misunderstanding :)
Peter talks about creating an automatic search for strategies based on an indicator, without yet considering the result of this experiment. And the guys want to immediately see the meaning of this action. That is why, when asked why, Peter says it is an experiment. Right?
It is possible to build such a system of searching for strategies. I have constructed similar through saving the results of a pass into a file, with further study of this file in subsequent passes. But in Quaternary. This is very convenient, no need to sit near a computer, charged for the whole day, and all the results are recorded.
Another thing is to look a little ahead. Is it possible to find a profitable strategy on standard indicators without wasting a lot of time? I marked a few ideal points to sell on the chart and threw some standard indicator.
What indicator values do we see in these points? There are only four ideal sell points in this timeframe. And how many similar indicator values? About 10-15. It's like that with any indicator. This is the problem.
I have an idea of how to automate the search and assembly of the strategy using the capabilities of the tester.
OK,
- it is necessary to pass data from previous passes to the current pass,
- make a block that will analyze the usefulness of the current result,
- create an option to skip useless passes in Oninit,
- maybe something else...
No problem - use Fourier approximation and extrapolation instead of zigzagging.
https://www.mql5.com/ru/forum/326818/page4#comment_14014309
https://www.mql5.com/ru/forum/325307/page2#comment_13709024
Peter, please understand that when you load an indicator, EA or even a script, all historical data is available to you via CopyRates or CopyTicks. And you don't need a tester for this, to pick parameters by brute force method.
Fourier is great to demonstrate the essence of your topic, because it essentially replaces N indicators, each of which contains 3 parameters(amplitude, period (frequency) and harmonic shift phase).
In a few milliseconds you can calculate for example 40 harmonics (which values are analogous to parameter values of any indicator), say for the last 10 years of history. And the grail for the history of the last 10 years is guaranteed.
The obtained harmonics can be used for future forecasting. But the problem is that such "optimization" on historical data does not work for effective prediction of the future, as well as with other indicators or their combination. On history it's grail, in real life it's a drain.
So, sorry of course, but then your topic is just another bullshit. This topic is only interesting for swindlers who try to rub their "miracle-advisor" into gullible potential buyers so that before buying the test it shows beautiful pictures and periodically updates "optimized" for history and for different symbols so that beautiful pictures don't turn into ugly truth.
there probably isn't enough page here to give you a plus.
I really like your work
+++
Make it a product, or they'll re-do it and get it rightThere's probably not enough page here to give you a plus.
I really like your work.
+++
Make it a product, or else the craftsmen will remake it and they'll get it rightThank you, Renat.
I just don't quite understand where the product can be.
It is an ordinary decomposition into harmonics, with the harmonics themselves displayed on the screen. The Fourier decomposition algorithm is not mine. I got it from the design bureau many years ago.
Fourier decomposition can only be used to good effect by a very few traders.
I wrote this code a couple of weeks ago in half an hour for one screenshot, which I needed for my studies.
If I post some code - it's not a pity and everyone may use it for whatever he/she wants. Sometimes I see on my marketplace that someone uses my idea. That only makes me feel good.
What indicator values do we see at these points? There are only four ideal sell points on this timeframe. And how many similar indicator values? About 10-15. And so with any indicator. This is the problem.
That's what I wrote about a couple of pages ago. Even if you find the indicators, the next challenge will be false positives, and the strategy will depend on the % of false positives and the profit/risk ratio of the trade.
By the way, if all the induks are fed to the input of the NS. And learn increasing the maximums of signals close to the ideal points. Isn't it the same thing? And hasn't it been done already?
imho, you won't find it, even on the non-ideal, in your opinion, ZigZag - you will be able to cover only partially with this selection of indicators, but you will not be able to cover completely all peaks of ZZ - neural networks do not count, they are for another purpose
i think that at least you have the structure of your future program, i can write it myself, but i'd be ready to use a ready-made program layout - i need it!
I agree about the zig-zag. For a trending strategy it's a perfect indicator on history.
I have the structure of the programme ;) I will implement the idea a little differently.