overlap of portfolio management and trading strategy

Discussion in 'Strategy Building' started by phattails, Jun 3, 2009.

  1. For those who have been trading a portfolio for a while, what do you think is the line that divides portfolio management and trading strategy.

    Sometimes the answer is obvious when there are few factors involved and scaling isn't an option.

    I am finding that as my strategies increase, my dimensionality becomes hazy because I feel like I have account for every dimension (strategy, asset, exogenous factors and especially multiple factors accounting for price shocks) .

    A very simple example would be trading an anomaly... so in this case you buy 1 lot and then sell when the anomaly is over. Now let's say you realize the anomaly is not black and white, but grey. But later you realize the grey areas have more to do with another anomally that you can't define.

    On one hand, I feel like I should make the agent dealing with the original model turn off the strategy at the grey parts because I don't feel like I've adequately accounted for this dimension. On the other hand I feel like I should let the portfolio deal with it. If this grey area were random, then it would be like dealing with any portfolio that didn't fully span the profit space and I would add more asset classes or dimensionality.

    So here's the question: Under the case I've specified, how do you manage those grey areas,?
  2. Not sure if this is related to the exact case you specifiy but, I can offer some of my experiences trading a variety of intraday systems. Basically, I believe that systems can be grouped into two distinct categories, directional or non-directional (trending or reverting markets). Some may have systems that fall into other catergories, but I do not and for simplicity's sake lets assume thats the type of market action we are trying to profit from.

    A lot can be learned from the first half hour of trading. Say the SPY is moving down strongly and doesn't make much of an effort at testing the opening highs, this would look to me like a directional down day, hence my directional systems would likely fire. Depending on the amount of capital automatically allocated to these short systems I would try to get a feeling of how strong the down move is going to be. Suppose I have a portfolio of 200 products that are high beta, if 50 products fire shorts then I know I ought to be careful if my mean-reversion systems (which will fire if the market goes down "too much") decides to start entering. If the down move is very strong, then the allocation for mean reversion should be minimal, unless the market starts showing signs of recovery in the midday or late afternoon. What the length of this post attempts to explain is discretionary element involved in intraday capital allocation. If the shorts are profitable I'll add to shorts, I don't want to add longs and vice versa... its the old saying: add to winners not losers.

    That's the extent of my portfolio level allocation. It has a lot to do with how the individual system P/L looks going into the day that decides where I start pushing more capital. I've been trying to get this hardcoded in software for a number of years but the problem of adequate simulation always arises... I've found it very difficult to simulate intermarket behavior on a portfolio P/L level. Still trying though...

  3. Let's start with the socratic method, so that we know we are answering the right questions.

    What is the purpose of a trading strategy? For me, a trading strategy attempts to map a unique segment of the potential profit space or space. It doesn't care about sharpe ratios or any other measure. The end goal of strategy development is to have mapped the most potential unique profit. Some people call this alpha, but my mind automatically thinks of a simple linear regression and therefore have a bias. I won't be calling it that because this is an excercise of thinking outside of the box, not scrambling in it.

    What is the purpose of a portfolio? A portfolio only exists for two reasons: we cannot fully map the potential profit space, and it accounts for dealing with transaction costs and limited liquidity. Realistically it attempts to be the human genome project, but with potential profit.

    These are answers I came up with and I hope someone can add some constructive criticism. Inevitably this will lead to more questions and down the path of wisdom, which I hope will start a concise fruitful dialogue.
  4. Purpose and meaning...

    Is this going to be a newbie fest about how model development and trading is some holy matrimony???

    Are you going to start placing your psychology and how emotions affect on top of systematic trading???
  5. Another newbie comment.

    You try to have different systems non-correlated and independent of each other for diversification. But yet, this guy adds useless discretion to make the systems correlated and kills the point of diversifying your portfolio.

    Did you start developing models due to some issues with discipline and the other BS about discretionary trading?

    Loser developer.
  6. Blah blah blah blah... god you must have a small d--k. Sorry about that one bud. Gotta play the hand you're dealt I guess.

    Oh BTW, some women just want to wear heels when they go out, don't take it personally if they laugh when you ask em' out, its just natural mate selection at work...

    Come to think of it, I'll help you out - you ever see beta chimps "swoop in" for sloppy seconds after the alpha male is done with the female? Watch some discovery channel when you get a chance - you might learn something to improve your odds.

  8. Here is another question related to my first post: To what extent can a portfolio map the p.p. space without the input of trading strategies? If we can express all factors in our trading strategies as factors in our portfolio, then we've got something.