Robotic/Auto-execution self-designed program trading

Discussion in 'Strategy Development' started by JT47319, Jun 8, 2003.

  1. JT47319


    Has anyone actually ever implemented a purely mechanical strategy that they designed/backtested and then sent to a purely robotic brokerage (ie strategy-runner) without any kind of discretionary input and let loose on the markets? Your strategy works, or it doesn't.

    Takes all human emotions like fear/greed out of the equation. Any thoughts? New wave of the future for the retail trader-- programmed trading?
  2. I trade fully automatic but never would trust a retail broker.

    Wave of the future or not, you will never be able to get rid of your fear and greed.

    Systematic doesn't mean:let it trade and forget about it. It means that you are abiding to your plan 100%, but what tells you that you will not suffer while the system is in a drawdown? What tells you that you will not change systems every two days because you think this system is not working the way it should. One way or the other, your emotions will pop up and it's not a bad thing if you're aware of them and know how to use them in a smart way. Reminds me of that book "Trading Psychology" by Steenbarger:

    If you have a red light saying your brakes are about dead in your car, would you rather have duck tape over it and never know when it pops up (therefore not having to deal with the fear when it does pop up when you're going 80mph), or would you rather see the light and take action accordingly? That's basically what your emotions are here for, they sense danger, whether you hit the brakes in frenzy and swerve all over the road and on the shoulder to try to stop your car or remain calm and focused and try to use other safer, maybe smarter ways to slow down is really up to you.

    Remember: the computer doesn't care about the money, you do.
  3. I would anticipate that by the end of the summer roughly 80% of my offices volume will come from fully automated trades where traders simply monitor the systems they have built.

    The only reason its not there now is; contrary to what some posters have had to say, getting an experienced programer to program an API or go through the api (I am just learning to code tradestation, so I have no idea how our guy does the real stuff) does not take a few days or a few weeks atleast for most of takes time to do it right, and to test and test and retest the system in 100 share lots over all types of days (long, short, trending, trendless, etc, etc.) long after you have done the backtesting over those same type of days and feel comfortable with the trading systems itself.

    Personally though I think anyone that trades spreads a lot has a good idea of just how much of that stuff is already automated today. The cancel and replace orders through ecn's on the few nasdaq related spreads (usai, pxlw, etc.) are enough to make a guy that is trying to manually do it throw his keyboard everyday at just missing one side of the thing again, and again, and again if an ecn was on the inside.

  4. I couldn't agree more. The automated portion of a system requires its own development cycle. Banging out some hack VB program in a couple of days is a recipe for disaster.

    At a minimum, we have a three month cycle where the first month is strictly paper trading (to test position allocation, general logistics), the second month places one share orders (tests interactions with the API, cancellations, etc), and then the third month is a limited scale real-time test where some of the desired positions are entered.
  5. maxpi


    Nobody anywhere thinks that unattended automated trading is a reality yet. There are aftermarket items that hook tradestation to IB. I'm working with that stuff as we speak but it's not up and running yet and when it is I'm going to be watching it all day.

  6. jeff520


    There's a lot of techniques that can benefit from automation. I use Neovest's IOG system and it allows me to do things I could never do if I had to enter trades manually (like taking advantage of short-term price aberations). I just wouldn't be able to get in and out (and update orders) fast enough without the computer's assitance.
  7. I'm working together with a programmer in order to build an automated trading system for NQ. It's going to be based on the 3 line price break method and uses tick bars. The program is java based and will work via IB's API.
  8. To answer the original question - YES. I manage a group of systems that I auto-execute at a few different brokers. Some are better than others. The system signals are generated from TradeStation files and execute through the broker's proprietary software directly to the client account. Execution is quick and accurate. Slippage compared to TS hypothetical performance reports is minimal (proof provided). If you're interested, check my profile for website link and contact through e-mail.
  9. CalTrader

    CalTrader Guest

    The answer is yes. The trick is that there needs to be an automated compliance system - rules based. That is the complicance system watches these automated systems: they have to go through the compliance engine in order to execute a trade: and if their position violates risk parameters then the engine requires them to exit their position. This serves as a self governing check on any aberrations of the modules that want to go and take on a position.

    If this is done correctly then an automated system - or an individual trader - will not be into a position that violates the risk parameters for a particular account.

    I often wonder how it is that ventures like LTCM blew up: I cant believe that they did not have these types of systems. Most likely they had poor management in place that allowed an inflated ego to roll the dice.
  10. cashonly

    cashonly Bright Trading, LLC


    I've done that.

    It CAN run totally automated, but it also has the ability for user overrides.

    There are 3 reasons for this:
    1. Problems can exist with... the quote feed, the execution platform, specialist doesn't respond, etc. Now this is rare, but it can happen. When it does, the program is designed to request a human takeover as it no longer has valid data to work with.
    2. I didn't but ALL my knowledge into the program (YET). There are a number of tweaks that I have on the list to add to the program and until I do I do them manually. They CAN be automated but have not been yet. And the program can run fine without them, but I find that it will run MORE profitably without them.
    3. All things CANNOT be incorporated into the program. For example, one time Greenspan announced a surprise rate cut and I saw it happen on CNBC. I marketed out of my shorts the instant I heard it and saved a lot over where the program would have gotten out because even with price action indicating the situation, the news beat the price in this instance and I could react quicker than the program.

    Think about it. Even though airplanes are so sophisticated they can fly the whole trip themselves, you wouldn't want to be in a totally automated one without a human to do a manual override, would you?

    So, while a strategy can be TOTALLY automated, I will always believe a human override ability needs to be available.

    #10     Jul 11, 2003