Page 1 of 502 12 LastLast
Results 1 to 10 of 20

Thread: A smart trading system?

  1. #1
    Hello everyone, first a word of introduction. I am new to this forum and to the forex trading as a whole. After reading a few books, playing a little using a demo account on and trying several different platforms I started my very first mini account together.
    Two days ago I found this forum and enjoyed the contents and the discussions very beneficial.

    I am currently studying bioinformatics and I thought of several systems that I might program to analyze the markets, as the field of bioinformatics is in reality dealing with the same problems as much of this specialized analysis - selective sign from noise using statistical procedures. I wanted to know your view whether it actually is worth my while (I am not always in terms of lots of gained gained money, but much more in relation to scientific interest (even if I found a nice scalping system that would provide me with just 10 pips a day I would be a millionaire in lt;5 years)).

    The systems I thought of trying to construct should be first of robust, i.e. they should work among as diverse market conditions as you can - and I also wish to ensure their evolvability. What I mean by this that those systems should change their egies as we go along. The implementation I think should be very computation-intensive and I am not certain that the issue isn't NP-complete. The key I thought of is implementing some kind of a genetic algorithm that the choice function could be measured most of all through robustness.

    Now another thought I thought of this might bring a lot to developing my system is mimicking pseudo-market ailments. Tell me what you think of the:
    A) Assessing the various parameters of a market and building pseudo-data with the same conditions employing random variations on the same theme.
    B) Perhaps searching for HMMs from the market information and seeking to reconstruct the information with those same HMMs.
    C) some kind of clever shuffling of the data.

    Do you think that using those parameters could still overfit the very best egy? Just because it'd find my parameters by which I constructed the market data?

    I would be very interested in hearing (reading) your comments to the ideas above.

  2. #2
    Quote Originally Posted by ;
    Oh btw - I forgot that I thought about the other modifiion of the system. How about integrating some news feed into the machine that is oriented on particular search phrases and combinations to comprehend the significance of it. Then modifying the egy so.
    This would definitely be a new spin on trading the news.

  3. #3
    Quote Originally Posted by ;
    I'm currently studying bioinformatics...

    Now another thought I thought of this might contribute a lot to creating my system is mimicking pseudo-market conditions. Tell me what you think of this:
    A) Assessing the different parameters of a market and building pseudo-data with the same conditions using random variations on precisely the exact same theme.
    B) Perhaps searching for HMMs from the market data and trying to rebuild the data with those same HMMs.
    C) some kind of smart shuffling of the information.
    welcome to the forum! Interesting to hear there are trading parallels in and bioinformatics. Person, if you realize that stuff, trading should be a breeze

    can you clarify what you mean by HMMs. What parameters are you thinking of measuring? It seems to me that you want to unleash an algorithm on the market to find some patterns. Is that what you are looking to do?

  4. #4
    Quote Originally Posted by ;
    can you describe what you mean by HMMs. What parameters are you considering quantifying? It sounds to me that you want to unleash an algorithm on the market to get some patterns. Is that what you are seeking to do?
    What I am looking mostly is to increase the accessible quantity of data. I am very much afraid that overfitting may happen if I use only the present data. (Recall that StocksCommodities article that random prices may also be very well back-tested and egies fitted to trade those curves).

    What I though t of by looking for patterns is for example:

    1) Define specific parameters manually that would explain the market. E.g. attempt to look for the stripes of time the market is trending (i.e. initial derivative averaged on X time points is higher/lower than Y brink ) of course define the variance of the tendency. Also check the second derivative of the price information (the rate of the price) and recieve some parameters for this also.
    Also would be interesting to see the amplitude of the market moves and the frequency of particular changes or waves.

    2) The other approach which I supposed by running Hidden Markov Models is learning the system parameters in the (almost) constant data (the price).

    After defining this set of parameters of workable information and avraging over many time points and throwing in a random element. I might have the ability to construct a data set that will have some similarity to the present markets, but not entirely like them and thus test my systems (or create them if I would like to utilize the GA) on a much more considerable number of data.

    P.s. another possible approach is shuffling the markets in order to produce the random information (possibly only time points inside the markets, and maybe even the markets as a whole after scaling (like mixing different currencies, however normalizing them to the dollar-adjusted pip-value).

  5. #5
    Do some of you know? It looks like the rate of change in price and the amount of time that a price is merging will give quite valuable insights about the significance of support and resistance. I really don't have a lot of programming experience but I'd like to build my own indiors in an intraday typing prices into excel everyday is getting old. It would be wonderful to have in automatic. Any suggestions?

  6. #6
    It is possible to download tick by tick data from:

    Nevertheless I do not know if I trust that their quotes entirely - they seem a bit too jumpy.

    Now about a platform - I do not know, but I was considering doing my work in MATLAB, it has big computing chances.

    About glow, there's a plugin, although I never worked with it would let you download automatically data and a free feed to excel through DDE.

    You might check it out at:

    The app is:
    (Be advised,however, that the site is in Russian, search the net, perhaps you'll discover other progr that might do the exact same thing).

    Hope it will help a little.

  7. #7
    Thanks ivenger. The Russian is a bit much for me to work through on this one. However, its great. Does anyone know of any other excel feeds such as ivenger mentioned?

  8. #8
    Quote Originally Posted by ;
    What I am looking mainly is to increase the accessible quantity of data. I am very much afraid that overfitting may happen if I use only the present data....
    This is a worthy cause, and youve revealed some true imagination in finding ways to get some random data.

    But why not simply buy some info? You can get tick data for the previous 10 years for under $100. I get mine from also it imports into TradeStation using a program called OwnData by TS Support. There are a bunch of different techniques to do it.

  9. #9
    Yes I could I downloaded the information for the previous 4 years at no cost by the website over ( The issue is that I am not certain that 10 years is sufficient.

  10. #10
    I am now inventing the required restrictions and the goals for the machine I wish to design. Please inform me if you believe that they are plausible.
    What I want in the system.

    A) I want to trade intraday. Thus the average trade must last anywhere from 15 minutes to five hours.
    B) The amount of successful trades ought to be 40% on average (try to grab weak movements ).
    C) The losses must be accordingly - the expectancy is a system that will make ~20per day on average on a mini lot(20 pips for USD/... pairs).
    D) The increase in position ought to be type of a pyramid (alhough I am not sure I want to restrict the machine in such a way. Let it pick by itself the increase/decrease in position in a trade.
    E) Maximal drawdown must be 20% of the equity. I'd also very similar to something that won't have over three consecutive losing days.
    F) The beginning capital ought to be very low.
    G) Time scale: The entrance as I see done by the machine ought to be based on a 5 min chart - without thinking about the condition of the one minutes for your buy (although do consider them to get your decision-making process). The 5-min chart is utilized just in order to make the system more powerful to my mistakes in implementing it).
    H) The machine will alone generate the exit signals that will adhere to the rules.
    I) The equity management also should be generated by the system

    What do you think?
    What I left out?

    Recall the system ought to be automatic - or just a mix of several uncorrelated systems.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
This website uses cookies
We use cookies to store session information to facilitate remembering your login information, to allow you to save website preferences, to personalise content and ads, to provide social media features and to analyse our traffic. We also share information about your use of our site with our social media, advertising and analytics partners more information