Which is easier - a steady 500 pips a month or just 20? - page 12
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 have no experience in building complex systems from atomic ones, but I have had ideas similar to yours for a long time. The optimizer in the strategy tester is not suitable for this purpose (it has no possibility to set an arbitrary user-defined target function). I am currently working on developing my own strategy tester with the ability to optimize.
P.S. Perhaps, the availability of such a possibility will parade us in MT5.
Yep. An optimizer with the possibility to set its target function is needed. Maybe, it is just not in demand. But for now. We need a good example.
Yep. An optimiser with the ability to set its own target function is needed. Maybe it's just not in demand. But that's for now. We need a good example.
Not only that. The optimizer is unsuitable even for optimization of the simplest systems. (get banned - not banned?) Although I think the tester itself is great.
A good example? Please! If you optimize any, even a relatively simple system with the target function Balance, you get what? Right. The genetic algorithm will select the ones with the highest balance from the variety of possible options. This will also include systems that have lost 300 trades and increased the deposit several times in a couple of trades. We don't need that. The same lopsidedness will occur if we select the remaining target functions - Profit factor, Maximal Drawdown etc.
Not to mention the problems that will arise if Mathemat decides to optimise his systems, which he mentioned above.
Not only that. The optimiser is unsuitable even for optimising the simplest systems. (will it get banned - not banned?) Although I think the tester itself is great.
A good example? Please! If you optimize any, even a relatively simple system with the target function Balance, you get what? Correct. The genetic algorithm will select those with the greatest balance from the variety of possible choices. This will also include systems that have lost 300 trades and increased the deposit several times in a couple of trades. We don't need that. Similar lopsidedness will result if we select the rest of the remaining target functions - Profit factor, Maximal Drawdown, etc.
Not to mention the problems that will arise if Mathemat decides to optimise its systems as he mentioned above.
I'm aware of all that. But I don't have enough knowledge for an optimizer. And about using my target function - I do my own, if necessary, with saving the result to a file in the standard optimizer. But as I see it, there is no target function in the existing optimizer. There is only the parameter by which the best (but not optimal) solution must be obtained. Though I don't know what is to be considered the best. But the fact that the resulting solution is not the optimal one is for sure.
I know about all this. But I don't have enough knowledge for the optimizer. And about using my target function - if necessary I do my own and save the result to a file in the standard optimiser. But as I see it, there is no target function in the existing optimizer. There is only the parameter by which the best (but not optimal) solution must be obtained. Though I don't know what is to be considered the best. But I'm sure that the resulting solution is not the optimal one.
Sorry, but you are wrong about absence of target function. Balance (and others) is the target function for the genetic algorithm that tries to maximize it. And those parameters which are in tab "Optimization" are just filters which sift out obtained results. Developers, please confirm. :)
Sorry, but you are wrong about the lack of a target function. Balance (and others) is the target function for the genetic algorithm, which tries to maximize it. And those parameters which are in tab "Optimization" are just filters which sift out obtained results. Developers, please confirm. :)
You can just optimize by balance, copy it into Excel, and use it to do ordering by any of your target functions with macro. But only on the basis of results of MT4 optimizer. Of course, the resulting options will not be entirely optimal in terms of the target function, but then you can additionally test within narrower bounds of optimized parameters.
It is crooked and a bit complicated, but possible :).
B.S. In a genetic algorithm, the target function not only eliminates the worst results, but also gives the best ones more resources to continue their research - to "continue their race".
You can simply optimise by balance, copy it into Excel, and use it to order any of your target functions with a macro. But only on the basis of results obtained by MT4 optimizer. Of course, the resulting options will not be entirely optimal from the viewpoint of the target function, but then you can additionally test within narrower bounds of optimized parameters.
It is crooked and a bit complicated, but it is possible :).
B.S. In a genetic algorithm, the target function not only eliminates the worst results, but also gives the best ones more resources to continue their research - to "continue their race".
It is not the target function that sifts out the worst results, it is the genetic algorithm. The target function is not part of the genetic algorithm. In nature, the target function is "survival". And it is not survival of individuals who only run fast, or only fly, etc. Survival is for individuals that possess the complex of these traits. In the MT optimizer you select those individuals, which, for instance, only run fast (Balance), or only those with bigger and stronger jaws (Profit factor). In life such freaks are not viable (real trade, or in other words, life).
P.S. Also, if I understand correctly, Mathemat is proposing to implement an entire eco-system of simultaneously living individuals of different species.
It is not the target function that sifts out the worst results, it is the genetic algorithm. The target function is not part of the genetic algorithm.
I don't know the specifics of GA implementation in MT4, but the target function is part of GA determining fitness of an individual, ranking them, increasing the probability of continuation of the species through mutation/crossbreeding/transfer to the next epoch.
I don't know the specifics of GA implementation in MT4, but the target function is part of the GA determining the fitness of an individual, ranking them, increasing the probability of continuation of the species through mutation/crossbreeding/transfer to the next epoch.
Genetic algorithms are in some ways as blind as evolution itself. GAs know nothing about the target function when they breed new individuals. Adaptability is tested independently of the GA, and the most adapted individual is returned to the GA, which will spawn new individuals from those parents. This is the main beauty of evolutionary methods of selection - complete ignorance of the developmental goals, in our context - of the goal function. And the fact that the GA itself and the target function are in the same software product, or code, does not mean that the target function is part of the GA.
Genetic algorithms are in some ways as blind as evolution itself. GAs know nothing of the target function when they breed new individuals. Adaptability is checked independently of the GA, and the most adapted individual is returned to the GA, which will spawn new individuals from those parents. This is the main beauty of evolutionary methods of selection - complete ignorance of the developmental goals, in our context - of the goal function. And the fact that the GA itself and the target function are in the same software product, or code, does not say that the target function is part of the GA.
There is nothing to argue about. You asked, "Are those parameters in the optimisation tab just filters that sift out the results". I replied no, they are not just filters that sift out the obtained results. A filter is either a yes or no filter. GA is implemented in a different way and based on the results of the evaluation function the individuals are ranked. GA also takes into account the value of the target function itself. For example, it compares best values of current and previous epochs as a stopping criterion.
There is nothing to argue about. You asked: "Are those parameters in the 'Optimisation' tab just filters that sift out the results? I replied no, they are not just filters that sift out the results obtained. A filter is either a yes or no filter. GA is implemented in a different way and based on the results of the evaluation function the individuals are ranked. GA also takes into account the value of the target function itself. For example, as a stopping criterion, it compares best values of the current and previous epochs.
I was not asking anything. I was asserting. I did not and will not convince anyone of anything. My methods work irrespective of any outside opinions.