The essence of optimisation - page 7
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 hate to burst your bubble, but optimisation is onlya POST-EFFECT, a finishing touch, nothing more. Looking for strategies in this way is confusing cause and effect, although this is a common mistake, among (algo)traders.
There's a saying about playing poker, "If you don't know who's a sucker at the table, it's you".
If you don't know better than optimization how to detect a pattern, you better start doing MLM.
Forward walk is a fallacy, if you cram the optimizer inside the strategy as an indicator, it's a real misunderstanding of what's going on at all.
Read Forward Testing, as an example of self-deception
The market is not that great, they just do not understand what's going on and they do not know what to do with it. Then they are surprised that one TS takes weeks to optimize. Shame on them.
Forward walk is a misconception, if you cram the optimiser inside the strategy as an indicator, it's a real misunderstanding of what's going on in general.
You are missing the point here. Walk forward is just the most effective way to bring the optimizer down to earth without losing money and time.
Well yes, sorry, I was categorical, it may be a little better than just in one iteration.
But as far as I understand, a person suggests using optimizer to manage adaptivity, of course, it is not really heresy, but it is a real mess. You do not understand how to monitor more directly the presence and dynamics of the pattern that is exploited by the strategy (algorithm).
If you are interested in optimization, try the genetic programming as I think it is the most global type of optimization, when not only the parameters for a given TS are searched but the TS itself
Alex_Bondar liked your article on habrahabr.ru/ "How I made a tester-optimizer to find profitable strategies on the exchange".
It seemed that on the contrary, you stand up for optimization as the main tool.
Now everything is mixed up, words, phrases, phrases. Some kind of single mental pillar!
Well yes, sorry, I was being categorical, it might be a bit better than just one iteration.
Well for the sake of comparison, the size of forward is rarely more than a third of the optimization interval, I'd even say a third is fat -- a fifth to a tenth.
With walk forward, you can look at several years' worth of stitching with an optimization interval of say six months.
Imho, it's very significant if we talk only and exclusively about optimization. All the rest is... The wrap-around and walk-forward will probably manage to "cheat".
Well for comparison's sake. the size of the forward rarely takes more than a third of the optimisation gap, I'd even say a third is fat -- a fifth to a tenth.
I always set it at 1/2.
See https://www.mql5.com/ru/code/2322 on the second screenshot.
And I advise others to do the same.
Robert Pardo's "Developing, Testing and Optimising Trading Systems for the Stock Trader". Who personally researched the material presented in it?
IMHO - the useful information is clearly present there.
On the fourth, the way people approach this question, although old, but, IMHO - the value has not lost!
I re-read it - it made me feel nostalgic and smile... :-)
PysSy a selection of articles from the fourth.
Alex_Bondar liked your article on habrahabr.ru/ "How I made a tester-optimizer to find profitable strategies on the exchange".
It seemed that on the contrary, you stand up for optimization as the main tool.
Now everything is mixed up, words, phrases, phrases. Some kind of unified mental pillar!
If you mean http://habrahabr.ru/post/209198/ I have nothing to do with it. I looked it up, it's some kind of marketplace bullshit, bordering on heresy.
Well for comparison's sake. the size of the forward rarely takes more than a third of the optimization gap, I'd even say a third is fat -- a fifth to a tenth.
With walk forward, you can look at a stitching over several years with an optimization interval of say half a year.
Imho, it's very significant if we're talking only and exclusively about optimization. All the rest is... The wrap and walk-forward will probably manage to "cheat".
Yes, walk-forward, reduces the likelihood of fitting and "smooths out" the results.
But in my experience auto-optimization did not work, in theory and before practice too it seemed that it is a super panacea, grail and so on. But it turned out to be rubbish.
Optimisation is about getting information about some lucky points of contact between your code and one of the manifestations of reality in the past. Expecting these hits to repeat in the future is like looking at a 5 year old child's autumn landscape drawing in winter to find the terrain in which it was made.
is not a true statement...
For a start - if you look closely at the graphs - even with the naked eye you can see repetition
there are patterns that are repeatable