How to optimise an advisor correctly - page 8
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
How to avoid it?
There is a parameter "minimum margin level %" in the tab "optimisation" - what does it mean - equity? and most importantly, how to set this limit correctly?
..genetic algorithm gives different results with continuous optimization...
More than once I was convinced that continuous optimization finds, well, not always the best variants, but it definitely finds variants that
"stole" the genetic algorithm from us. My personal tragedy is that my "codes" have been continuously optimized for months.
So I will die with a suspicion that somewhere in the basket there is an undiscovered grail...
to rider
If the balance crashes at the end of testing, I consider the drawdown unacceptable and reject the variant.
IMHO, this is a specific denigration of the strategy.
Why go straight to marriage? If you put six of these advisors on different pairs, it'll be like in the movies, hippy end.
I'll vouch for the "end," but what do you mean by "hippie"? A fun dog with songs?
to granit77
Where do our ideas, our notions of what a truly worthwhile advisor should be, come from?
- do they just come from somewhere or is it a sober calculation?
And if it is not a belief, but a calculation, then again, from what kind of beliefs does it come?
--
I.e., in addition to TS, we have a model of how an Expert Advisor works, under which we search for and select our TS.
but who can prove that this ideal model, which fits TS, is not an obsession?))
Otherwise, how many TPs are rejected because they don't fit the ideal model and how many are rejected because of real drawbacks?
for example:
well what makes us think that the stop should not be more than 10% of the depo?
If it is fair to say that the stop should not exceed 10% of the deposit, at what lot?
maybe without emotions the best MM would be: deposit 10k, lot 0.1?
You're really going for the basics, brother...
I have no answers, I'm not a philosopher, I can only say that on the basic question (advisor suitability) my mind is in a mess.
The calculation may only be based on individual critical parameters, which, again, everyone picks for himself. I'm not a very good example,
Since I am lazy and uneducated, but my preferences will be outlined, especially as you have called an ideal model an obsession.
My obsession is a TS with one open order and minimum possible drawdown (3-5%), working with indicators which I
I understand, i.e. when visually testing it does not give the impression that it "thinks for itself". As a consequence, it needs
only in rough optimization, under pair and TF and persistently (at least a little) profitable on the whole history. For optimization 0.1 lot, 2K deposit, after optimization
lot can be increased up to 0.5 with no risk of MC, then TC starts "falling" at the least profitable areas. MM - percentage of balance, entered last.
In general, I do not understand the stop on the deposit; the levels of all colours - only for self-reassurance; everyone draws waves in his imagination; the statistics is only as an
and Mathematics - for indicators and Mathemat, etc. All in all, savagery and ignorance, go to bed...
2 granit77: That's how it goes. I try and try, like carrying some decent-looking maths, but then comes gray-haired practical person granit77 - and says that all this does not need hell, from the evil it is, all this maths. However my obsessions are not too different from yours. So mathematics is not for changing obsessions.
Well, who am I trying to do it for, if not for people like you? But, no, I know, first of all, for myself, strange as it may seem. When I started writing my "masterpieces" I didn't know where they would lead. I made discoveries for myself.
I hope I didn't break anyone's associative line?
The way to optimise depends on the Tc underlying the EA itself.
let's assume - there are some generalized rules - methods of out-of-sample runs as well as selection of optimal values
more I would probably like to point out the following! https://championship.mql5.com/2012/ru/news ( Testing and Optimising Experts )
---
alexander (bettor) has a very interesting method for value selection
when fitting and optimising his EA for the 2007 championship, I used my own method of parameter averaging
---
of course every TS has its own indicators, the system may have completely different criteria
I achieved a good ratio of profitable trades to unprofitable ones with the correct alternation
something like 1 1 1 1 1 0 1 1 1 1 1 0 1 1 1 1 1 0 1 1 1 1 0 1 0 1 1 1 1 1
losses 1 0 0 0 0 0 1 - killed the system because it was possible to increase the lot after losing - i.e. MM was included in the system as a component
Alexander's system has a 50/50 ratio of profitable and loss-making trades, but profits are growing and losses are nipped in the bud
--
that is why in different TS the desire to increase some indicators at the expense of others is unreasonable
---
but the methodology of sampling and running and averaging parameters may be similar
granit77 писал (а) >>
Убеждался не раз, что сплошная оптимизация находит, ну, не всегда лучшие варианты, но однозначно находит варианты, которые
"украл" у нас генетический алгоритм. Моя личная трагедия в том, что мои "коды" всплошную оптимизируются месяцами.
Так я и помру с подозрением, что где-то в корзине гниет ненайденный грааль..
to rider
Если в конце тестирования рушится баланс, то я считаю просадку недопустимой и без сожаления отбраковываю вариант.
ИМХО, это конкретный порочащий стратегию признак.
Try to divide your EA into several, sensibly of course. The one I have now is optimising across the entire sample and across all ticks, asking for 1448 hours. I didn't like it, so I divided it by four according to exit rules. Already 60 :)
To granit77. Screenshots still can not download, so look at the attachment. :( .......... there NZDUSD240 Out of Sample for two months (if I remember correctly), so, if testing/optimisation around the 95th trade is stopped, the balance will "crumble" - and then? And you're suggesting this should be thrown in the bin?
Yes, changeable lot, it's not a martingale or MM, the trade is 0.1 lot - and that's total position management by direction - an integral part of the EA.
In this thread or not, I don't know, but somewhere there was something about recovery factor: Profit/Maximum drawdown should be at least 30 or something else.
Here's an example:
test period in months/profit/maximum drawdown/recovery factor
2/500/500/1
4/1000/500/2
......
12/3000/500/6
24/6000/500/10 etc.
What are you supposed to do with it?
I hope I didn't break anyone's associative line?
Not at the moment. Nothing and nowhere is off the charts :)