Multiple System Allocation

Discussion in 'Strategy Development' started by trend456, Aug 18, 2003.

  1. I have read and heard so many things on creating and applying money management. There's one thing that I haven't had a good answer when trading multiple systems or trading for example 10 different markets.

    Nobody hasn't really touched this deeply. I think this is an important topic. Most of the topics are based on one system.

    Do you guys have a systematic approach in terms of locating assets with multiple systems?

    I wonder how huge managed funds allocate assets to different traders or systems.

    Good Trade!!!

  2. I understand this can best be studied via portfolio level backtesting, using 3rd party TS RINA plugins, Trading Recipes, Behold!, or more sophisticated simulation platforms like smartquant in .net. I have only worked with TS so far, but for example, you might optimize a portfolio of system allocations for ROA, which would give you maximum linearity in your equity curve.

    I've also just begun studying this topic, and from the little I know, it looks like inverse correlation and equity linearity are the primary goals in allocating capital in a portfolio of systems.

    I currently trade one system on two symbols that are almost 100% correlated with only slight variations, and am looking for more diverse systems to balance things out.
  3. Hey vanilla,

    What's up!!!

    Actually.... I'm just getting into this topic deeply since I have multiple systems running. I think this is a great topic since it can be applied to so many things. I have optimized a portfolio of systems but found that it wasn't the best way to go.

    But... thanks for your comment.... Something to think about. I have some systems that are negative correlated but I still can't get the answer.

    I'm also wondering how much portfolio heat people put up trading multiple systems.

    Good trade!!

  4. I think when trading multiple systems on the same symbol, it's important to trade different time frames. The bigger the difference in the time frame for each system, the less correlated the results will be for each.

    For example, using a system with 10 minute bars might get chopped to death on low range days, while a 30 minute bar system will see you catch the overall trend a little better, yeilding good results.

    The only exception to this is if you use one trend following system and one trend fading system. One will always outperform the other depending on the underlying volatility, except you get a much smoother equity curve.

    I know a lot of hedge funds operate in this fashion. During low volatility periods, multiple systems tend to negate each other, resulting in little to no profit, but minimising the posibility of a major drawdown. When the volitility returns, more overall trend following signals are issued, while less trend fading signals are issued resulting in a directional bias for the portfolio of systems. The total positions will be mostly long, or mostly short when the market is streaming in one direction.

  5. Runningbear,

    What is the maximum portfolio heat one would allow trading multiple systems? What do u use to calculate portfolio heat? OptimalF??


    Good Trade.

  6. Hrmm... it looks like most people haven't touched this too deeply too. Or they just don't or can't do it.

    There are key components to this.

    1. Correlation. Simple eh?

    2. Capital allocation. OK, this is the hard part. But this all depends on the portfolio of systems you have.

    On a single system there are systems that work better under % capital risk, some work better under a static $value, some better with constant lots and raising it every few $1000 and so forth... This goes the same with portfolio.

    Just like creating a system, it's just how much tools in the shed you have. These come from studying and experience, also some instinctive recognition of system reports, and hypothetical trades made...

    Just my 2 cents.
  7. Dude,

    I know that part.

    I'm trying to systematize this whole process.


  8. heya trend bud

    excellent suggestion runningbear, i forgot about multiple timeframes.

    i actually did trade my system on two timeframes for a few weeks. predictably, the high frequency system looked outstanding on paper, but after charting my fills against the simulation, the gap gradually widened over time to a point where it looked dubious at best. Maybe if I slow it down a little more, but still faster than the long term system, I can achieve the correllation benefits we're talking about.

    Also, don't completely discount the benefits of optimizing the allocation variables in an entire portfolio of systems for ROA. I've never tried it on a portfolio, but this seems to me like the most rational way to maximize linearity accross systems. I've had a lot of success optimizing single strategy exits for ROA, and see no reason why the concept would apply any differently to allocations within a portfolio. I'll let you know when I get portfolio level testing set up. For now, these are just the speculations of a beginner.
  9. vikana

    vikana Moderator

  10. Any suggestions on Multiple markets...? I have multiple systems on multiple markets. I'm sure some of you trade multiple stocks with different systems. How have u been handling it?

    Good Trade

    #10     Aug 18, 2003