ATS development

Discussion in 'Automated Trading' started by hfmanager44, Jun 25, 2011.

  1. Well firstly I would like to introduce myself to you ET members reading this article. I am a nineteen year old finance/mathematics student with two years of trading equities and currencies. I have recently come to the conclusion that I need to build an automated trading system so I can attend class while "investing". However, I have been researching several avenues, but have little idea on where to proceed.

    I've written up the GUI, features, strategy, etc. for the program I'm trying to create. I was wondering if I could get your opinions on the vastness of this topic and if you could direct me to the right resources.

    Some of my ideas include:
    Neural Network+ Genetic Algorithm + Simulated Annealing
    Automated Stock, bond, currency, option analysis
    value-at-risk calculations
    automated trading (obviously)
    looking at C++ coding
    1 day prediction of stock, bond, etc. price fluctuation
    Detailed charts with numerous indicators.

    Is my list too vast or impossible?
    Could one please give me a list of recommended readings, resources, code examples, other needed features?

    I would greatly appreciate any and all help!
    Thanks in advance!! :)
  2. May I give you some advice?

    Many traders appear to be misguided in trying to implement automated trading via APIs et al rather than performing basic data analysis first. Those entering the market are so driven to trade NOW and immediate automation of the actual trading, they under emphasize the data collection and data analysis.

    Your first task should be to collect good quality historical data in the areas that you want to trade in.

    After this, spend 90% of your time focused on data normalization and de-noising. Books covering signal analysis and behavioral economics would likely be beneficial.

    Just do this in your spare time for a couple of years -- ignore the actual trading activities.

    Good preparation will set you up for success.

    Good luck.
  3. Yes, I understand what you are saying and completely agree with the idea that many people want to automate trading instead of having to put the time into actually understanding basic investment principles.

    However, I've been researching heavily in the quantitative finance and financial engineering concepts to create this "super" program that can be the Soros, Durkenmiller, Bacon, etc. of hedge fund elite. My thing is that I don't believe in value-investing, to me, if you're only bringing in 10% or staying with the S&P it isn't good enough. Soros brought into investing the concept of reflexivitiy(?), which in my opinion is far superior to the trade strategy used by the Tiger funds, Buffett, etc.

    Perhaps you are right in that I am prematurely postulating about such an acute subject matter without having the vast proper foundation. I would hope that you are wrong for my pride's sake lol, but other than that what else do you think would be a good route?

    I mean, for me, I have had fairly good results in Forex and equities with a 50% combined ROR in a year so I figured, "Hey, if I can do this, why not just build a program that can be more accurate and faster in analyzing data, picking stocks, hedging them with options, and watching for slight triggers in volume that could cascade into something greater."

    I've also been studying the GMAT, CFA, Series 6,7,63 testing materials to understand as many aspects of finance as possible.

    Gracias Amigo :)
    Your knowledge and insight is greatly appreciated.
  4. Understanding Zero Sum Game math and randomness...
    Is so focking central to trading...
    That is you're not a savant in this area...
    Like you were not the Math Genius in your Grade 5 class...
    And have not been obsessed with gaming, etc all your life...
    Just go do something else like start a rock n' roll band.

    In addition...
    The biggest problem with ET "rocket scientists"...
    Is that they fantasize about HFT, etc...
    While lacking any meaningful trading expertise.

    You guys are like building a Brain Surgery Robot in your garage...
    Without being a doctor or actually having any experience in surgery...
    Put that in your crack pipe... and take a big hit.

    Expertise in ANY field requires 10,000 hours of dedication...
    Some will say obsession.

    No one here will ever "automate" their way to success...
    Without becoming an experienced, profitable trader first...
    And understanding what works and what will never work in a million years.
  5. Pick one, small, stupid trading concept you think will work, then build that using what you know while learning what you don't know. Don't waste time building useless libraries without an objective in mind.

    Odds are, your first few iterations will be garbage. Just be sure to learn from what went wrong with the garbage on your path to greatness.
  6. rosy2


    do you have a strategy already?
    writing the infrastructure and creating a profitable trade are two different things.

    I can sell you a basic infrastructure if you like, PM me
  7. Not what I was getting at. If you're going to develop an automated system, especially using NN's you'll need clean, detrended, normalized data or the NN or other logic network will give non-normal results.

    To do this you first need to understand the data, especially seasonality, and this applies to any market.

    "Other than that, Mrs. Lincoln, how was the play?" First you need quality materials (data), then I'd suggest you understand all the tools in your toolbox (GARCH, ARCH, ARIMA, ARMA, variance and heteroskedacity, wavelets, FFT's). Read every post by Acrary twice. Since you're using NN's, read "Neural, Novel & Hybrid Algorithms for Time Series Prediction" at least twice. Understand the curse of dimensionality and how it relates to NN's. Understand how Acrary's approach, though simple, addresses this.

    I know you're in a rush, but spend 90% of your time in preparation and 10% in execution. Take years lining up your shot and when you fire, fire with grenades.
  8. Occam


    I agree that trading usually takes a lot of learning, but unforunately for most people who put in those 10,000 hours you suggest, they'll still end up with no profitable strategy, without having learned any useful/fun/fulfilling skills whatsoever. In this respect, trading is really quite different from virtually any other profiession.
  9. cloudy


    Try out some ideas on MT4 's C like source .mq4. There are tons of free samples of indicators and even EA's out there. there' a reference manual on .mq4 somewhere..
  10. there is a good blog that covers this topic as well. the author of the blog posts a basic trading system and asks for readers inputs on how to improve the system. the author then will post the backtested results on the blog. its a pretty cool idea. nothing great has come from it yet as it just began but a pretty cool concept i think.

    the address is
    #10     Jun 27, 2011