Where is the line between fitting and actual patterns? - page 13
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
I'll jump in, if you don't mind, Vitaly.
That's why Pardo's book was written. There are a lot of criteria, it's not that simple...
But the most important criterion is known: the FN must be somewhere within a certain "stability area", i.e. the area in which small changes in parameters do not lead to a drastic change in the system profile.
How to choose a single final set (FN) of parameter values? What are the selection criteria?
If you have found the optimal parameters for each major, it means that there should be correlation between the parameters, at least a decrease in the profitability of the TS for one instrument - the first sign that it is necessary to adjust the parameters again
SZZY: I have not tried to make a simple TS, that would work for any major when adjusted - I think I will start doing it soon, just thinking aloud
Before voicing your thoughts, I would like to know what regularities you have in mind?
Laws of the found sets of optimisation parameters? Or?
For example, here's an interesting pattern that worked well a few years ago, I'm sure works now perhaps with different settings, but the approach itself is the same... - http://forex.kbpauk.ru/showflat.php/Cat/0/Number/25750/an/0/page/0#Post25750 - a pattern of hourly movements of the euro. I know that people have built entire TS on this approach to the market.
Igor, look, there was an optimization with three parameters, each having 10 values (agree, this is very modest).
A total of 1,000 passes.
How to select a single final set (FN) of parameter values? What are the selection criteria?
Do you know the answer to this question?
Without answering it, there is no point in going any further!
And here are the variants that need to be tested on forwards: 1. by the principle of maximum presence of a parameter in the sample, 2. by the average value of the parameter out of, say, 500 flatbeds. 3. According to the best optimization pass (by profit, by min drawdown, etc.) (depending on your choice)). The optimization period itself must be representative, i.e. regardless of the type and type of the system the period must include different market phases, and the number of deals must be sane - from 200, 300. Forward testing should cover 10-25% of the optimization period. After each optimization period, test the parameters on the forward on these 3 selection criteria, after the 7th, 8th and 9th forward you come to a certain criterion and the last optimization period (10th) results in the selection of parameters by the previously detected criterion and go to the real account (in time, again, not more than 25% of the optimization period). That's the end of the cycle. If you like the results on the real account you start "practicing" again. :-))) (This is to the word "opting", the rest is not a joke).
Read more here - https://www.mql5.com/ru/forum/107064 and in R. Pardo's "Testing,..." - third time I write this, and the questions are the same on the subject.
And here already variants to be tested on forwards: 1. by principle of maximal presence of parameter in sample, 2. by average value of parameter out of, say, 500 flat. 3. By the best optimization pass (by profit, by min drawdown, etc. (as selected by you). (depending on your choice)). The optimization period itself must be representative, i.e. regardless of the type and type of the system the period must include different market phases, and the number of deals must be sane - from 200, 300. Forward testing should cover 10-25% of the optimization period. After each optimization period, test the parameters on the forward on these 3 selection criteria, after the 7th, 8th and 9th forward you come to a certain criterion and the last optimization period (10th) results in the selection of parameters by the previously detected criterion and go to the real account (in time, again, not more than 25% of the optimization period). That's the end of the cycle. If you like the results on the real account you start "practicing" again. :-))) (This is to the word "opting", the rest is not a joke).
Read more here - https://www.mql5.com/ru/forum/107064 and R. Pardo's "Testing,..." - third time I've written and the questions are the same.
1) I agree with much, much is negotiable.
2) This thread has long been a favourite of mine
3) Can we jointly try to do something, as suggested here?
1) I agree with many things, many things are negotiable.
2) I have long had this thread in my favorites.
3) Shall we jointly try something as suggested here?
1. Yes, Vitaly, I agree, many things are negotiable...
3. The holidays are over... Off to my main job... Let's see... I'll look into it "as suggested here?"...
I'll jump in, if you don't mind, Vitaly.
That's why Pardo's book was written. There's a lot of criteria, it's not that simple...
But the most important criterion is known: FN must be somewhere inside some "stability area", i.e. the area in which small changes in parameters do not lead to a drastic change in system profitability.
I'll get into it... Well, what are you, Alexei? I always feel like you're out there... Really. )
...............
On topic.
Pardo is great, but no answers.
Some "areas", some "stability" .....
.........................
In the picture, in the area under the "?
What areas of sustainability in the preceding optimization period can tell us that almost 99% on OOS will end in a loss. (below the red line)?
The problem with Pardo's perception is that the book is looking for some kind of guarantee. There are no guarantees. All the many checks are only necessary conditions, but not sufficient.
In other words, Pardo says: "Guys, what I've written to you here are the obligatory signs of any profitable system: if the system is profitable, it must have these signs. But these attributes are not sufficient to make sure the system is profitable.
OK, off to bed.
The same as our attentiveness has a lot in common.
Without boring the reader with pictures - I will say that from "pipsing" to "long-building" we ourselves program the result.
By testing with our hands! A well-known fact of fatalism.
And meta(mega)quotas just help.
;)
This is self-defeating.
In the process of developing the TS, the lot must be constant.
(Again, slyly.... )
??? I'm sorry, owe who? Who said that? What book does it say?
My lot, I change it as I please.