Algorithm Optimisation Championship. - page 106
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
Of course, trading narrows the field of application of this algorithm.
I think ...
Can you without blah-blah-blah give a clear and understandable example - application of "optimization algorithms" in trading?
If you don't know such examples - just say - "I don't know what kind of trading tasks need fast optimization algorithms".
Can you give a clear and understandable example without blah-blah-blah-blah, the application of "optimization algorithms" in trading?
If you don't know such examples, just say "I don't know what kind of trading tasks require fast optimization algorithms".
Great.
It turns out that in order to find a compromise with the participants and to organise the competition properly - you just need to rack your brains a bit...
About the notorious universality you talk so much about - I've come to the conclusion that it doesn't always produce the best results.
1. The universality of a solution is always relative, because the solution is limited to the specifics of the problem domain - and hence - the solution is never absolutely universal. When expanding the problem domain, a "universal" solution will always fail. It will have to be reworked.
2. No universality appears from scratch, but is a consequence of a long process of development, generalization of problems and adaptation of the solution. Hence, non-universal solution is the first step towards universal solution.
3. The universality of the solution does not mean the effectiveness of the solution. I think that these two notions are not directly connected and do not depend on each other.
Striving for universality makes one adapt the solution to a widening range of problems, which of course may reduce the effectiveness of the solution in each particular case.
My algorithm for text mining is universal enough for text mining, and can absolutely precisely identify any string in the minimum number of times I accessed the FF. Perhaps its further development, may lead to finding maxima of unknown analytic functions. But will it remain effective in this case? I'm not sure.
And so, to understand how we can make a universal algorithm, we need to generalize the range of problems and understand the general mechanism of their solution.
First, let's summarize the parameters.
The main parameters with which the algorithm works to find the maximal value of the function and the text key:
1. The number of parameters passed to the FF.
2. The range of the values of the parameters passed to the FF.
3. step (minimal difference between the values).
4. Value received from the FF.
In the absence of more basic parameters, the solution, even without any extra effort, may turn out to be quite universal...
The search mechanism in these two types of these problems can be generalized, which I will try to do.
You stubbornly ignore my advice. I have vast experience in the field of optimization algorithms, though it sounds immodest, and at least listen to my advice, right?
There is a huge class of problems which cannot be solved analytically. And only a very small part of all problems can be solved analytically. But you are focused exactly on analytics, leaving all the other tasks aside.
I never claimed that universal solutions are always better, on the contrary, it is quite clear that highly specialized solutions will always give better results. But when speaking of universality as applied to optimization algorithms, we mean the capability of such algorithms to solve any tasks, including analytical ones, i.e. absolutely any tasks. So your approach covers, say, 10% of possible problems, while mine covers 100%.
We are not only theoreticians here, we are first of all practitioners. We trade on the financial markets and for us there is no analytical formula of the market, so we need to use solutions that were not originally designed for analytical solutions. This is what universality is all about as well when it comes to researching market patterns. I have shown you several times the examples in the face of MT's in-house optimizer and Alglib in the codebase, which are universal optimization algorithms that do not require knowledge of the FF formula.
Everything has its flip side of the coin. In case of universal optimization algorithms this is a necessary compromise which leads to lower accuracy of the solution, but it's just impossible to solve trader's problems in any other way. And when I said that I had found a solution of composing FF with the known maximum for the referee, you would have to ask "what is the downside of such a solution? Firstly, it does not allow mutual influence of all parameters in FF, which is a simplification, and secondly, there is a way to cheat in the competition, but I do not think that someone will be able to use this opportunity, because it requires a lot of time, which will not be between championship steps. And this "first" is the most important thing, but I had to do it for the sake of a fair demand of participants to be able to compare algorithms with the real maximum.
Can you read carefully? It clearly says - tweaking strategy parameters to get the highest profitability in the tester.
What does this have to do with the fast and efficient algorithm for identification of an extremum?
Speed and efficiency are simply properties of the algorithm. They are not mandatory.
The extremum sought during testing is the maximum profitability over the historical interval for the trading strategy under test.
When an extremum is found, the tester stores values of strategy parameters that led to it and shows them to the user.
This is called "fitting".
For some reason you stubbornly ignore my advice. I have tremendous experience in the field of optimization algorithms, although this sounds immodest, and at least listen to my advice, right?
There is a huge class of problems which cannot be solved analytically. And only a very small part of all problems can be solved analytically. But you are focused exactly on analytics, leaving all the other tasks aside.
I never claimed that universal solutions are always better, on the contrary, it is quite clear that highly specialized solutions will always give better results. But when we speak of universality as applied to optimization algorithms, we mean the capability of such algorithms to solve any tasks, including analytical ones, i.e. absolutely any tasks. So your approach covers, say, 10% of possible tasks, while mine covers 100%.
We are not only theoreticians here, we are first of all practitioners. We trade on the financial markets and for us there is no analytical formula of the market, so we need to use solutions that were not originally designed for analytical solutions. This is what universality is all about as well when it comes to researching market patterns. I have shown you several times the examples in the face of MT's in-house optimizer and Alglib in the codebase, which are universal optimization algorithms that do not require knowledge of the FF formula.
Everything has its flip side of the coin. In case of universal optimization algorithms this is a necessary compromise which leads to lower accuracy of the solution, but it's just impossible to solve trader's problems in any other way. And when I said that I had found a solution of composing FF with the known maximum for referee, you would have to ask "what is the downside of such a solution? Firstly, it does not allow mutual influence of all parameters in FF, which is a simplification, and secondly, there is a way to cheat in the competition, but I do not think that someone will be able to take advantage of such an opportunity, because it requires a lot of time, which will not be between championship steps. And this "first" is the most important thing, but I had to do it for the sake of a fair demand of participants to be able to compare algorithms with the real maximum.
Speed and efficiency are simply properties of the algorithm. They are not mandatory.
The extremum sought during testing is the maximum profitability over the historical interval for the trading strategy being tested.
When this extremum is found, the tester saves the values of the strategy parameters that led to it and shows them to the user.
This is called "fitting".
I see -- in general, your understanding of the need for "optimization algorithms" in trading is not even close. Everything is at the level of general phrases and invented abstraction.
I think you have a problem with understanding elementary things.
Once again In trading the optimization algorithm is needed to fit the values of trading strategy's parameters to a certain period of time in the past in order to use them in the future.
I think that you have a problem with understanding elementary things.
Once again: In trading the optimization algorithm is needed to fit the values of trading strategy's parameters to a certain period of the past time in order to use them in the future.
Let me rephrase the question. The subject is the development of an "optimization algorithm".
Can you popularly explain what an "optimization algorithm" is and give clear examples of its use in trading.
What is interesting to hear:
-- you will develop some functionality based on the results of this "championship". Give an example of how this functionality can be used in trading.
Let me rephrase the question. The subject is the development of an "optimization algorithm".
Can you popularly explain what an "optimization algorithm" is and give clear examples of its use in trading.
What is interesting to hear:
-- you will develop some functionality based on the results of this "championship". Give an example of how this functionality can be used in trading.
Imagine there is no tester in MT.
You yourself need to test your strategy on a selected interval of chart history of some instrument.
Your strategy has 5 parameters, whose values affect the profitability of your strategy, on the selected test section.
You need to determine which values of these parameters will give the highest profitability of your strategy at this section of the chart.
There are so many possible parameter values that if you manually check each of them, it will take you years to find the ones that give the highest profitability to your strategy.
You begin to develop an algorithm for optimizing your parameters that will calculate their best values for the tested period faster than you.
Why? Because you have decided that history will repeat itself and a similar period will happen again in the future.
You believe that then you will apply the best values of the parameters found by your algorithm, and become rich!
The entire practical use of the algorithm in trading may not be worth a penny, but if you believe that the future will be similar to the past, and if it happens, then such an algorithm will bring you a fortune.
The Championship is an opportunity to check your skills in practice and compare them with those of other Expert Advisors. In my opinion, the practical benefit of the Championship is greater than the practical benefit of using the optimization algorithm in trading, because I do not believe in such a precise repetition of history, when once adjusted parameters will give the same result as during the testing period.