Optimisation vs Fitting in the story - page 3
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
Well, he's usually a man of his word and deed
This is not the case, and he didn't promise anything. It is such a mythical goal to strive for - when optimising try to calculate rather than over calculate.
Is there an example of this block in CodeBase?
Of course, it is not a universal block. It is different for each strategy. This block is 99% of the strategy's success and complexity. It is, in essence, the AI. This is what Max is working on.
Consider, for example, the simplest strategy with one parameter:
Of course, we can find an "optimum" МА period on any time frame where the strategy will be profitable on this period.
Further on, if we do not change this MA period, it will be a guaranteed drawdown - it is only a matter of time.
For real self-optimization when the МА period is constantly revised depending on the market situation, we need this block of AI which, in brief, solves the task of pattern recognition. There are a googol of options here.
For example, the simplest one is to create a 6-dimensional space of points. Each point has 6 coordinates:
Thus 6-dimensional clouds are formed and when analyzed(it is a longer history) they can be "forecasted" for the required current MA period.
The 3rd coordinate is useful for determining a flat/trend.
This very large amount of calculation occurs only once per one pass of history data. Further with each new tick (bar) points are only added to this space.
Forward optimization with clear (and the same for all forward runs) selection criteria (they should be) for selection of optimization variants for this very forward run.) In this case, it is optimization, not fitting. If it works, I have not seen such Expert Advisors that would have clear criteria for selecting optimization variants and that would successfully pass this forward optimization)).
What is the fundamental difference between these concepts?
History matching is a set of values for the parameters of an Expert Advisor which give the best trading results on a certain historical timeframe of price movement.
Now try to define optimization.
Are they the same?
Forward optimization with clear (and the same for all forward runs) selection criteria (they should be) for selection of optimization variants for this very forward run. In this case, it is optimization, not fitting. If it works, I haven't seen such Expert Advisors that have precise criteria for selecting optimization options and successfully pass this forward optimization).
Here is the forward test for example. What is the conclusion?
Optimisation and fitting are the same thing.
Here's a forward test for an example. What is the conclusion?
no. for conclusions you need either a walk-froward test or a backtest with a built-in auto-optimiser.
Optimisation is a broader concept. History fitting is one of the methods of optimisation (direct calculation of function values). In essence, an EA is a function described in a peculiar way. MT tester calculates its values.
Optimization is a search for an extremum of a function by one or several parameters (that's what I remember from university). The tester offers you a set of values but it does not perform optimization as per the definition - you do that selecting from the offered values.
PS
It seems that one question was asked, but another question was implied: how to assess whether the extremum chosen is local (for a given time interval/sample) or global.
Unfortunately, I got a C in statistics with a stretch :( No help.