Automated Trading for a Living

Discussion in 'Automated Trading' started by Allistah, Apr 30, 2013.

  1. Allistah


    Hi all --

    I need some help. I'm reaching out in hopes I can get some good advice from those that are successful with automated trading. I've read a lot of posts which had a lot of good information.

    My goal: Quit my job and trade full time via automated trading

    First thing, I am not planning on quitting my job any time soon. I know that this will take a long time and I am willing to put in the time. I've already put in nine years and I'm willing to put in another nine or more to get this right.

    Some history on myself: I have been trading for nine years as I said. I have lost a lot of money and made a lot of money. I want to keep it going in the right direction now. I have been writing strategies for about seven years now. I am a programmer and have a lot of experience coding strategies. I have only done manual backtesting which has been difficult with my current trading platform. I have not been able to do forward testing. I have been victim to over curve fitting. It is indeed deceptive and can suck you in making you think you're a genius.

    I am currently moving to Tradestation so that I can make use of its back and forward testing tools. I am not concerned with making strategies. I have made quite a few and don't really know if any of them really work because I haven't done the right testing on them.

    Most of my strategies are intra-day. I may look to do some based off a daily chart at some point but I'm starting with intra-day. What I am looking for are tips on what I need to do next to get to my goal. Here are some of my questions:

    1) How long do I need to do backtesting?

    2) How much forward testing?

    3) How would I know if a strategy is good enough to start to take live with real money? I would like to back and forward test, then paper trade it, then go live with it if everything is good. The problem is, I just don't know how to say when each step is "good".

    4) Are there any books that are highly recommended for this stuff?

    5) I've heard that you really need to be running multiple strategies on multiple symbols to be successful. Is that true?

    6) Is there anything else that you would like to share that you wished someone would have told you before you started with this?

    7) Are you trading full time using automated strategies? Is there anyone really doing this for a living?

    I've been at this long enough and I know that making a lot of money up front is ridiculous and won't happen. I am ready to start on this automated path while I work my full time job and some day when the day comes, I want to quit my current full time job.

    Thanks for any help you might have for me.
  2. I've run an automated trading system for about a year now. I am now taking in external capital and plan to do this full-time as soon as the AUM becomes sufficiently large. Check my trading journal "Quantum Leap" on this forum for performance. To answer your questions:
    1. The amount of time is not necessarily the most important thing. From a statistical viewpoint, you first of all need enough data points to determine whether you have achieved statistically significant results. Secondly, you need to test your strategies through bullish, bearish and consolidating market phases, which usually requires several years of data. The more the better though.
    2. In my experience a few months of forward testing (paper trading) and compare results with out of sample back testing again to rectify discrepancies.
    3. If you have back tested, forward tested, and paper traded, that should be good enough.
    4. My recommendation would be "Algorithmic Trading" by Ernest Chan
    5. False! At least I haven't needed to.
    6. You need a lot of blood, sweat and tears!
    7. Not yet - in my case it's now a matter of bringing in capital though.
  3. Allistah


    How often are you re-optimizing your strategies? I'm not sure if you're able to share this or not, but it is relative to the amount of backtest and forward testing you're doing. If you backtest for 4 months, then forward test for 1 month, you might re-optimize every month. It might be minutes. Is there some sort of rule of thumb ratio for back testing/forward testing and how often you want to re-optimize?

    Also, do your strategies run 24/7 or are they only running during market hours? There are a lot of opportunities in the night hours but obviously you need to sleep.

    Also have you ever let the strategy trade unattended? Was just curious as to how long. I mean, if you need to go in the kitchen and make lunch, thats probably not a big deal but if you let it run overnight while you sleep, thats something completely different.

    Thanks for taking the time to reply. I really appreciate it.
  4. In my experience, you should avoid strategies which requires re-optimization, as this is always done in retrospect, meaning that there's a risk that you will often be "out-of-sync" with the prevalent market sentiment.

    My current strategy only trades during opening hours for technical reasons. But from practical reasons it's also more convenient as you have natural interruptions during which you can do fine-tuning, updates etc.

    I never spend 100% of my time monitoring the system. If you paper trade for a while you can ensure all (or most) ins and outs of the algo have been tested. In the beginning of the live trading I might spend a few days of more or less constant monitoring, but nowadays I only check it momentarily, as it automatically starts and stops.

    It even keeps track of the contract roll-over days throughout the year! I have also set up a web server so that I can monitor on any mobile device, even while sipping margaritas on the beach... :)

    Good luck!
  5. Allistah


    Avoiding strategies that don't need to be optimized.. Hrm, does that mean you don't use any indicators - ever? It is all based on price action alone? Either that or the indicators you're using are all using values that are static?

    Almost all of my strategies use some form of an indicator one way or another. So is this going to cause me a lot of problems?
  6. No no, you misinterpreted me - I wrote "re-optimization", not "optimization", that's a big difference!
  7. Be aware that you are talking about a serious undertaking here and a lot if staged against you. I am in a similar situation, except I am a little further, have 2 people working full time on strat development and am lucky enough to not have to work anymore.

    Anyhow, a lot of the frameworks you will deal with - retail software - is utter crap written by people at the lower echolong of programming capabilities never using them. I Talk of non-repeatable backtests (run them 3 times, get 3 results - different ones), I talk of funny crashes and dead terribly slow backtests.

    Either you manage to work with this, or you are in a LOT of work to do your own framework (which is the direction we went). The state of retail software is purely insulting, sadly.
  8. Jusr to give an idea what I am talking about. We run our own infra.

    Now, we had some days with - issues. Mostly we found some data bugs, so now we reexported about a terabyte of market data stream and rerun some optimizations.

    Try that in ninja: 55 optimizations must be run over the weekend. That is 55 market/strat combos. You have 8 computers for that.

    "Ninja" (and other retail): Someone sits and watches.

    Good: I just watch a website showing the the queue of jobs. The first one right now has 3263 tasks done, 54 in progress and 1233 scheduled. Every task is 1 week of optimization, max 128 parameter combos (if there are more, they are split).

    Difference? Noone has to watch the whole thing and start the next optimization.

    This is partially what I mean when I say the retail stuff is - ah - challenging, from a professional and usability point of view.
  9. gmst


    Good points. Can you post a pic of your application, would like to see how your desktop with this application looks.
  10. No, plus - this is a web app. I mean, not the applicaiton to write strats (that is visual studio), the control and initial analysis (just some playing around) panel for the grid system. PLus there is some seriously confidential data on it - strategy name (haha) but mostly legally tricky stuff (employee names - the guy who "owns" a job, i.e. has scheduled it). It is late here, let it run overnight, then tomorrow I make a screenshot and redact out some stuff.
    #10     May 2, 2013