![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
"All the hard-earned goods."
As I do....... only, Mr. Reshetov - I am no longer your branch - no need to fall out...... for the approach and idea thanks for the first time already :)
There are two ready-made Expert Advisors 1.0 - positions by direction are accumulated, 1.1 - one position to the opposite signal.
After selecting a strategy the Expert Advisors are practically ready both for demo and for real - "cutting off unnecessary things", that's all.
A few explanations for *.set - 100 quid at 0.01 lot - limitation on the drawdown ...... further you can and should add 1000000 and the corresponding percentage, so that the amount does not greatly vary, depending on the deposit
Step 1.
- generate a strategy on the interval 2005.01.01 - 2007.12.20
- perform forwards (the script for Exel is enclosed) 6-6-3 (months)
- we choose the one with comparable profit on these periods; the number of deals is also comparable
Stage 2 (strategy and everything else is already chosen)
- delete everything unnecessary :)
- add "crutches and supports" in the form of TP, SL, Tral, etc. I have a variant with the fractal trawl - I like it better
- we optimize the pick on the same period, as in phase 1
- forwards - the principle is the same
Stage 3
- optimization (otherwise - adjustment), but over the entire history
- demo......
Next, the fog :))))
PS:
- Someone will say that routine - but try to automate(!?)
- Attached here are both codes and sets, I can not find the Exel code, if interested - I will post it later
- And, please, I did not go to work today, my computer is weak at home - if someone has something smart, put..... it all on all timeframes and symbols (phase 1) is calculated in one day, the analysis-selection - it takes a lot longer, of course
- note that everything is done with an "extra" decimal point.
- The code is written "in bits and pieces", so I apologize to the authors for its anonymity because........ myself do not pretend to authorship in any way.
- Don't push too hard, please :))
Forward loses its meaning when used repeatedly. By repeatedly running the forward and filtering by its results, you just implicitly fit the entire history section: test+forward. The probability of the adjustment is directly proportional to the number of forward runs and to the ratio of test/forward period length.
All those strategy generators are just a fitting with a large number of degrees of freedom, and the tests are of no help. There will always be some variants on history that will exceed the robust strategies by any criteria and that is what you will search for. It's simple combinatorics.
... In history there will always be fits that will outperform robust strategies ...
Another important criterion is the robustness of the resulting strategies.
Avals, you can define the robustness characteristics. What are the parameters you need to know and evaluate in order to classify a strategy as robust?
Another important criterion is the robustness of the resulting strategies.
Avals, can you outline the characteristics of robustness. What parameters of a strategy need to be known and evaluated to classify it as robust?
This is unformalizable. There are partial methods of estimation, but they are not universal. It depends on the type of systems and the patterns used.
Much can be understood from the width and "quality" of the optimal zone or the invariance of the optimal zone at different test periods. Multicurrency, but again not for all systems. Robustness needs to be evaluated both for the system as a whole and for its individual parts, which is not always a trivial task and requires specific tests. There are more or less common heuristic methods of robustness estimation but anyway each system requires separate approach. imho.
Forward loses its meaning when used repeatedly. By repeatedly running the forward and filtering by its results, you just implicitly fit the entire history section: test+forward. The probability of the adjustment is directly proportional to the number of forward runs and to the ratio of test/forward period length.
All those strategy generators are just a fitting with a large number of degrees of freedom, and the tests are of no help. There will always be variants on history that will outperform robust strategies by any criteria and that is what you will search for. It's simple combinatorics.
100% agree.... there are variants that all add up and even on demo it goes into minus.... it's not a bug in the tester, it's a harsh reality )))
In general, all the most interesting thing on optimization turns out - its period, period of forwards, period of EA without overoptimization..... how to calculate these parameters?
your option, plz.....
your option, plz.....
a variant of what?
if you're looking for robust strategies, i don't know a universal one, see previous post.
I do not know the universal criterion for robust strategies, but it is practically the same exact criterion that has not gone far from what beginners are looking for.)
Robustness is first of all a simple and logical trading idea, not a search for ways to transform prices + search for transformation parameters. It's like looking for a needle in a haystack with a pitchfork ;)
a variant of what?
If you are looking for robust strategies, I don't know a universal one, see the previous post.
The universal criterion for robust strategies is practically the same grail that has not gone far from what beginners are looking for)))
Robustness is first of all a simple and logical trading idea, not a search for ways to transform prices + search for transformation parameters.
corrected the previous post a bit.....
This is basically the following question: you can create an unprecedented great amount of Expert Advisors. But how and on what period of time should I optimize them, and by what parameters should I evaluate their reliability - the question?
Going no further than this is key......
The universal criterion for finding robust strategies is practically the same grail that beginners look for))) ...
It's like looking for a needle in a haystack with a pitchfork ;)And still, there are successful traders who manage to find them, and then they put away their pitchforks and use shovels. :)
My thoughts concerning the search of robustness criteria are the following:
1. If a strategy gives results not only on the main instrument (where it was generated), but also on a couple of other instruments, then it is a sign of robustness.
2. If a strategy passes back and forward tests, that is also a sign.
3. Robustness has a period and it can be "expired". Therefore it does not make sense to look for a strategy with robustness at all times. It's enough for N bars ahead. :)
100% agree.... there are variants when it all adds up and even on demo it goes into minus.... It's not a bug in the tester, it's a harsh reality )))
In general, all the most interesting is the optimization - its period, the period of forwards, the period of EA without overoptimization..... how to calculate these parameters?
your option, plz.....
I don't think these parameters should be calculated directly. A particular TS with a set of some parameters is one of final implementations. Its robustness by itself cannot be estimated separately. The TS is based on trading ideas which must be checked for robustness, not their separate implementations. But one can check the robustness of a trading idea only through the set of its final realizations. A robust trading idea must have a lot of high-profit end realizations for each history interval, for many instruments, and these end realizations must be in a fairly narrow range/set of ranges. Moreover: for each single piece of history the best profitable variants (profit, profit factor) must fall into the optimal zone during optimization. But this is all a heuristic like many other methods. The goal is the same - to find optimal methods to minimize the probability of fitting.
corrected the previous post a little bit.....
Basically, the question is this: you can create an unprecedented number of experts, but how and over what period to optimize them, and by what parameters to evaluate their reliability - the QUESTION?
You should not go further than this - this is the key.......
The question is rather rhetorical. Clearly, the optimization should be done in that period, in which the strategy will be used, i.e. in the future.
The problem of the Buridan donkey who famously starved to death between two different but appetising and fragrant hay bales without being able to solve the problem of which bale should be used to starve the worm. Quite often there are also buridan traders who believe that there are supposedly unambiguous answers to questions in the context of market instability, and therefore, in their opinion, one should first get a specific categorical answer and then get down to business.