Dynamic Portfoloi Selection

Discussion in 'Risk Management' started by dima777, Aug 2, 2008.

  1. dima777


    I wonder if anyone is using the method which would dynamically pick and drop stocks from the basket of those traded based on their performance as opposed to a static combination of stocks traded as a portfolio.
  2. theres a function in Amibroker that will rank stocks. It hasn't really improved the returns when I tested it based on any criteria such as Rate of Change, Rate of Change divided by the Stdev..... etc...

    I assume if you design a model which tracks behavior of individual stocks you can arb that in a long/short fashion at the portfolio level. You could also model the position size of each position in conjunction with the allocation to maximize the returns or specify the desired returns versus drawdown characteristic.
  3. dima777


    thank you for your reply..i am thinking to rank the stocks mainly based on how the system operates on them....if they perform well paper trading with the current system they are taken into the portfolio...when the reverse happens and they start to perform badly they are "cast down" to the paper trading level..
  4. lately


    This is what adaptive models are. They look for conditions where a strategy is profitable. If the strategy starts becoming unprofitable, the model can adapt and change.

    One way to define this:

    The task comes down to finding the 'best' ruleset of conditions that defines the domain of profitability for a given strategy. In an adaptive model, this can be evaluated based on activation relative to exit. If the character of the market you were exploiting goes away (falls past threshold) the model will use a strategy that is profitable under those conditions (or nothing at all, in your example).
  5. dima777


    thank you for your answer...can you recommend any book on this topic? I assume the classic portfolio theory books do not cover this topic...correct me if i am wrong...
  6. dima-

    The short answer is that the only good books on trading you will ever find are books on trading psychology.

    No one is going to give away anything that works. And what was relevant in the 90s momentum Markets (What most books you see these days are written about are mostly useless). Besides, people with strategies that are extremely profit guard them, as I'm sure you would too. Most likely, they give away strategies that have poor risk reward characteristics or have lost their edge entirely.

    BTW, have you ever in your life while reading any trading book or article in stocks and commodities or active trader seen anything that was backtested as a profitable long/short system that made more than 15% annually on a consistent basis? The answer is no no and no. All those posted systems are garbage. Thats the first thing I learned. There is no free ride. You have to do the training yourself.

    Learn a retail type software like Amibroker or Wealthlab for a year, or buy some data and create something in java to run backtests yourself.

    For portfolio theory- its all very abstract and the latest stuff isn't even able to work in the real world. Check out:


    Here again, you'll have to find a strategy yourself that is capable of being adaptive on multiple stocks and adjusts the size and allocation accordingly.
  7. dima777


    thank you for your detailed reply....i will now read the Markowitz's book on portfolio theory...thanks for those links....i am starting to fell that i am going to need to create this model from scratch..
  8. dima777


    one question - you think that 15% annually on a consistent basis is a worthwhile goal for strategy development?