users' questions

Тема в разделе "General questions", создана пользователем Dmitri, 16 Июль 2014.

  1. Dmitri

    Dmitri Бывалый Команда форума Администратор

    Регистрация:
    7 Ноябрь 2013
    Сообщения:
    4.136
    Симпатии:
    3.196
    Баллы:
    435
    Пол:
    Мужской
    Я на Forex с:
    2008
    Hi.
    First, my Russian is terrible. I learned it when I was a child but without practice...
    I'm a developer as you. When I want to evaluate new ideas I always look for ways of fast prototyping. I'm used to this from Java, C, SQL whatever.

    - MetaTrader + MQL really is a pain in the ass. Too many bugs and/or undocumented "features" mess up everything, especially because every new broker brings in new discoveries/surprises. After 4 years I have a sophisticated MQL framework, inc. timezone management, history management, backoffice interface etc. etc. Still a pain...
    - StrategyQuant is of no use for anything more advanced.
    - ForexTester is great but you need to develop a prototyping framework beforehand.

    If you don't mind:
    What approach or software do you use for testing/fast prototyping of new ideas? I'm looking for something simple and fast and discover that I spent way too much time on maintaining MQL libraries to keep compatibility with every new terminal realease. Which piece of software or approach did I miss?

    Best regards,
    Peter
     
  2. loopsider

    loopsider Бывалый Команда форума Модератор

    Регистрация:
    13 Ноябрь 2013
    Сообщения:
    4.735
    Симпатии:
    4.384
    Баллы:
    435
    Пол:
    Мужской
    We are in fact more traders than developers. Software means something only if it helps you to trade. The web is full of pretty codes that are completely useless when it comes to the actual trading. From this perspective, the question of framework is secondary. The primary question is what you are going to trade and how.

    MQL and the stategy tester have many limitations of course, but within those limitations mql is by far the fastest way to check you ideas ("fastest" meaning not the execution time of course, but the time before you will get answer to your question). Of course if your ideas require heavy mathematics or minimal latency, you would have to abandon mql or in fact MT altogether. Everything depends on the problem you try to solve, I do not think one can develop a universal framework and be happy forever after.

    We generally keep things as simple as possible. mql and MT strategy tester + tick data and 99% modelling backtesting stuff. In some cases it is easier to write a code (in any programming language) that reads ticks data directly from a file and processes them.
     
    bellduke, chickendinner и Dmitri нравится это.
  3. Scott

    Scott Новичок - ARGOLab.net -

    Регистрация:
    13 Август 2014
    Сообщения:
    5
    Симпатии:
    1
    Баллы:
    3
    Пол:
    Мужской
    Question is about back testing.

    I understand that in live trading the EA runs on each of the pairs in a 15M chart. What I'm not understanding is that the back testing is all on the 1M period. Isn't this going to generate far more trades than the 15M period which was advised to load the EA on? I'm guessing that I'm not understanding something.

    I would think that since it runs live on the 15M chart that you would backtest the same period. If you do back test on 1M how does the EA know to use the 15M period for trades if the tester is using 1M chart???

    Scott
     
    Dmitri нравится это.
  4. loopsider

    loopsider Бывалый Команда форума Модератор

    Регистрация:
    13 Ноябрь 2013
    Сообщения:
    4.735
    Симпатии:
    4.384
    Баллы:
    435
    Пол:
    Мужской
    Trading of EA does not depend on the timeframe, so you can backtest it on any timeframe.
    In real trading, it is recommended to attach EA to M15 or to keep M15 chart open.
     
  5. Scott

    Scott Новичок - ARGOLab.net -

    Регистрация:
    13 Август 2014
    Сообщения:
    5
    Симпатии:
    1
    Баллы:
    3
    Пол:
    Мужской
    Ok, but why do your back test on the site use 1M periods but you recommend 15M for trading?

    I don't understand what you mean by "or keep M15 chart open" Please explain this
     
  6. loopsider

    loopsider Бывалый Команда форума Модератор

    Регистрация:
    13 Ноябрь 2013
    Сообщения:
    4.735
    Симпатии:
    4.384
    Баллы:
    435
    Пол:
    Мужской
    EA uses M15 bar quotes quotes. In tester, it does not matter which timeframe you chose, results are the same. In actual trading, it is better to use M15 chart, in order to ensure that EA gets M15 quotes without disturbance.
     
  7. Dmitri

    Dmitri Бывалый Команда форума Администратор

    Регистрация:
    7 Ноябрь 2013
    Сообщения:
    4.136
    Симпатии:
    3.196
    Баллы:
    435
    Пол:
    Мужской
    Я на Forex с:
    2008
    Message Body:
     
  8. James Robertson

    James Robertson Постоялец Аргонавт

    Регистрация:
    18 Февраль 2015
    Сообщения:
    114
    Симпатии:
    95
    Баллы:
    55
    Я на Forex с:
    5
    I run a number of Argolab EAs already, and am trying to decide what to buy next: Splitter or Strike.

    Splitter is proven and I like the idea of running both licenses with different set files, maybe R and W, for diversification.

    Strike is newer and looks sophisticated, also more expensive to run multiple accounts.

    I would love to hear what everyone here thinks about the choice! ;)
     
  9. loopsider

    loopsider Бывалый Команда форума Модератор

    Регистрация:
    13 Ноябрь 2013
    Сообщения:
    4.735
    Симпатии:
    4.384
    Баллы:
    435
    Пол:
    Мужской
    On the present market, Splitter looks more stable, I would say.
     
    James Robertson нравится это.
  10. Dmitri

    Dmitri Бывалый Команда форума Администратор

    Регистрация:
    7 Ноябрь 2013
    Сообщения:
    4.136
    Симпатии:
    3.196
    Баллы:
    435
    Пол:
    Мужской
    Я на Forex с:
    2008
    Polling, activation schemes Пожалуйста, войдите или зарегистрируйтесь для просмотра ссылок
     
  11. James Robertson

    James Robertson Постоялец Аргонавт

    Регистрация:
    18 Февраль 2015
    Сообщения:
    114
    Симпатии:
    95
    Баллы:
    55
    Я на Forex с:
    5
    An idea for future development (or if you know an EA like this exists, please let me know!):-
    A very useful tool would be an EA that can resolve a locked position aiming to minimise losses (not necessarily remove them completely). Basic idea is to close positions by scaling out (partially closing) of the long positions when the upper boundary of a dynamic channel is reached (Bollinger bands or similar to FFSingle), and scaling out of the short positions when the lower boundary is reached.
    The partial close amount would need to be small enough to allow for multiple closes in the same direction, without the net position growing too large. Perhaps the EA could set a max of, say, 3 closes in one direction before waiting for a close in the other direction.
    Optionally, useful features would be:
    - a measure of whether the primary up/downtrend that caused the locked position has calmed enough to attempt recovery
    - a running p/l, so that, for example, if the first 3 partial closes are all long and so the drawdown (unrealised + realised) at this point has increased, then the EA keeps count and targets a larger profit / smaller loss from subsequent closes.
    - a "wait" function to ensure that successive closes don't happen too close together
    - ability to measure the primary trend in order to capture more of the profit from that direction than the other; eg if there's a strong uptrend, then perhaps after a partial close of a short trade (so the basket is net long) then the notional TP (ie the level at which the basket goes back to neutral, or reverses net direction) is set further than it would be when the basket is net short.
    I'm sure there's something in this that will work, and I know you guys can do the mathematics (and coding) !
    Smile15289
     
    Dmitri, loopsider и bellduke нравится это.
  12. loopsider

    loopsider Бывалый Команда форума Модератор

    Регистрация:
    13 Ноябрь 2013
    Сообщения:
    4.735
    Симпатии:
    4.384
    Баллы:
    435
    Пол:
    Мужской
    Thanks for suggestions.
    Resolving locked position automatically is a very difficult topic. Generally, it is hard enough to get any stable profit from the market. Resolving locks, you need not only get some profits but profits large enough to compensate your (usually, rather large) locked drawdown.
    The main problem with your suggestions is that this logic will - in most cases - close positions with loss. And who needs an EA which brings only losses? Nobody. People do not like losses...
     

.

Поделиться этой страницей

translate