Algorithm Optimisation Championship. - page 118
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
Andrey Dik:
accuracy and the number of calls to the FF are two evaluation criteria, with accuracy being 3 times preferable.
Respected topicstarter, remember how recently you ignored the issue of accuracy of the result when searching for a maximum that is reliably unknown.
In one of your posts, you made it clear that a reliable maximum was impossible to find, and therefore the "accuracy" criterion falls away.
Then you decided to make the maximum you were looking for reliably known.
After that, there is a new attitude that 3 times the accuracy is preferable to the number of hits.
Just consider that when you are told that the championship rules were not originally thought out, this is by no means without merit.
Then answer the question: Why is accuracy 3 times more valuable than the number of FF calls!
Where do you fit in, do you even know what we are talking about in this thread?
You have taken the phrase out of context and as usual do not understand.
There are two options: to strive for accuracy with an unlimited number of calls to the FF and a limited number of calls to the FF. Optimization algorithms are used where there is limited time to optimize, i.e. the second option means that the quality of the algorithm is judged by the accuracy with which the maximum is found with a limited number of FFs. The second option is more natural for the application conditions of the optimization algorithm.
And this is what you consider a worthy argument for determining a winner?
The answer is simple:
1. If an algorithm CANNOT find an extremum with a given accuracy, it has no place in the Championship;
2. Considering point 1, only algorithms that SEARCH an extremum with a given accuracy will participate in determining the winner;
3. no ranking in terms of accuracy. Accuracy is given by a range;
4. the winner is determined by the number of times the FF is accessed.
I'll leave at this point, it's 2 a.m.
Fantasies of a theorist far removed from the subject.
An optimization algorithm is not obliged to find a maximum, because there can be very tricky functions and the purpose of the optimization algorithm is not to do exact arithmetic. Any even a crank-created optimization algorithm is likely to find the maximum, if given enough time. So point 1 is of no importance.
2) Is there a new championship organiser?
3. See point 2.
4. See my previous posts. The optimization algorithm is not looking for an extremum, but for a maximum, it doesn't know when to stop.
Jooooooooooooooooo, where and in what have you sinned so much? Maybe you should go to church, confess, or something like that...
Andrew, look, every Mr. can't get past the topic without coming in and kicking you.
From the point of view of the practice of using AO in trading, which consists primarily in adjusting the values of the tested strategy's parameters and searching for variants of values that maximize profitability, the number of calls is much more important than precision.
Testing time and CPU load depend on the number of calls, while accuracy determines the difference in profitability in a cent range.
So that is the practice.
Jooooooooooooooooo, where and in what have you sinned so much? Maybe you should go to church, confess, or something like that...
Andrew, look, every Mr. can't get past the topic without coming in and kicking you.
From the point of view of the practice of using AO in trading, which consists primarily in adjusting the values of the tested strategy's parameters and searching for variants of values that maximize profitability, the number of calls is much more important than precision.
Testing time and CPU load depend on the number of calls, while accuracy determines the difference in profitability in a cent range.
So that is the practice.
So set a limit on hits, what's the problem? So in the championship, there is a limit, less - please, but more - ni ni!
In my opinion, this is not the right way to prioritise EAs.
In trading, AO is a tool used for approximate (and not absolutely accurate) calculation of the strategy profitability, and the acceptable error may be within a dollar.
However, if you strive to achieve maximum accuracy (which, by the way, has the same far-fetched meaning as the belief that the values found in the future will bring the mountain of gold), you can excessively overuse computer resources and your time.
Practice requires us to be rational and use working tools effectively.
... For example, if the optimisation step is 0.01, then the values obtained by the competitors must differ from each other by the second digit after the decimal point. ...