Automating the search for strategies. - page 4
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 assume that strategies can be transmitted via HTTP protocol, MQL has a possibility to receive strategies in this way.
I want to make everything fully automated, search for strategies, make portfolios of strategies, transfer to the Expert Advisor, etc.
Part of the system in MQL is 90% ready, working with a lot of strategies (position control, risks, error handling, etc.).
There is still a lot of work to be done.
Well impressive in general (although there is a lot of incomprehensible stuff). Congratulations. I would love to work with such a device. In general, this direction is promising and will definitely be developed. My only personal wish is to be able to generate the result of generation in the form of a series of Expert Advisors. It would be more convenient for me to test them in the autotester as a list. The history of generation is also important, because often a trader's capricious imagination leads him to another dead end and then it is important to quickly return to the starting point without losses.
This is not a problem, you can write down a whole list of strategies in the Expert Advisor and then test each strategy separately. Just take into account that in the MQL4 strategy tester there is no possibility to use a lot of symbols, in the beginning strategies are fully checked for correctness of execution (code correctness, presence of symbols).
Strategies will be written to a file, then the Expert Advisor will read and execute them.
Every time I load variants into the autotester I think about it. Here's what I'm thinking
1. The strategy generator should work on the principle of evolutionary tree from simple to complex.
2. The variants should be immediately checked on the volking-forward and eliminated
3. Functions should be prepared manually, and the generator should work out only the variants of their interaction, i.e. create interdependencies.
By the way, in the English thread I met a mention of some Bulgarian software with elements of something like this. But since it was on MT4, I was not interested in it.
And here is another German one, also on MT4 http://darwins-fx-tools.com/.
I know such a programme - StrategyQuant (only for MT4). It costs a lot, but it somehow finds strategies by itself. The problem is that these strategies lose a lot in the forward test. For example, you can easily make an Expert Advisor on EMA-crossover strategy, which will give excellent backtest results, but completely fail in the fronttest. This programme will produce a thousand of similar strategies. To create a really profitable Expert Advisor, you need a lot of work to screen the used indicators and to choose the optimisation criteria.
Therefore, in my understanding, a strategy generator is not just a constructor where you build a strategy from ready-made blocks, editing these blocks and creating new relationships if necessary.
It should be integrated with an autotester like volking-forward. Its purpose is to automate routine work - so the most routine work is selecting indicators, checking different types of dependencies and testing.
And ideally - selection by the principle of evolution according to a given algorithm through an autotester. I remember how I selected the types of zig-zags involved in an Expert Advisor for a month! And what did the process consist of? I just took an indicator, prescribed it, ran it in the tester, looked at the forwards, compared them, took another one, prescribed it, ran it in the forward and so on. I had to leave a series of these Expert Advisors to be tested overnight. I did something for the second time, something I skipped.
So this ruin kills creativity, you gradually forget that other solutions are possible and accept intermediate ones.
I generally believe that strategy search can be fully automated, the only question is the complexity of writing a programme for this case and the availability of computing resources.
In general, I believe that strategy search can be fully automated, the only question is the complexity of writing a programme for it and the availability of computing resources.