Multi-Strategy Money Management

Discussion in 'Risk Management' started by ezbentley, Mar 20, 2009.

  1. Suppose I have an account with 1 million dollars(I don't) and I want to trade 5 fully mechanical strategies on equities for diversification purpose. Ideally, the returns of the strategies should have low or negative correlation to maximize the benefit of diversification.

    However, different strategies can have different exposure, and it's possible that sometimes the positions required by the strategies can exceed the available buying power of the account. Strategy A may generate between 0~20 signals each day, with 5% of account equity per trade. Strategy B may generate 10 signals per day, with 10% of account per trade. There is no problem if they never generate signals on the same day, but when they do, they will fight for available buying power.

    Of course a dumb solution is just to reduce the percentage allocate to each trade per each strategy so they add up to 100%. But I doubt that's the most efficient way to manage a portfolio. Also we cannot just randomly allocate available buying power to strategy A or B, since that will cause the real-time performance to deviate from backtesting each strategy individually.

    I am not looking for a quick answer here. Can anyone recommend any book or reading on position management of a multi-strategy portfolio? I am quite interested in money management and am willing to do some hard reading.

  2. It might be overkill, but Active Management by Grinold and Kahn and Quant Equity Portfolio Management by Qian Hua and Sorensen both talk about similar issues. Another book that is more basic and has a short section on portfolio optimization (with cookbook excel examples) is Financial Modeling by Benninga.

    I think your problem fits into the generic framework in these books but with some tweaks. You would probably have to treat each system or signalled trade as a stock or alpha factor, and forecast the expected returns, volatilities and correlations. You come up with constraints and optimize the portfolio. Every time you get a signal, update the forecast, reoptimize the target portfolio, and decide whether it is worth rebalancing the real portfolio to get there (considering transaction costs and uncertainty) or just letting it sit.

    But I think it is probably overkill, if your forecasts are not that good and not much different from each other you would doing a lot of work for not much improvement over some reasonable rules of thumb, especially with just 5 systems. One alternative would be to simulate performance of several different rules of thumb with monte carlo analysis.

    So my prediction is still that this stuff is a lot of work for little gain in your scenario, but let us know if you figure out anything different. Good luck!
  3. MGJ


    In addition to reading, I would suggest you do some experimenting too. Purchase some high end backtesting software, program it with your trading strategies, and then try out some risk-allocation ideas in backtesting. Let the books be a starting point in your efforts to accumulate ideas, and let the test software help you decide which ideas you like best (i.e. which ideas work the best in testing). After trying the stuff you find in books, you'll be ready to make your own tweaks and insert your own little (or big) changes and new ideas.

    The three backtesters that I'm sure can do the job, are Power ST, Mechanica, and Trading Blox. The other programs aimed at retail customers cannot. One I'm not quite sure about is AmiBroker: can it simulate 5 different systems using 5 different portfolios, simultaneously trading out of the same account? I don't think so, but could be mistaken.
  4. there are numerous threads on ET that discuss this concept. Search around for threads by TSGannGalt and Acrary and they should be somewhere in there.

    Also, here is some software that should help out too.

  5. Thanks for the replies. They are all very helpful.

    I searched for "multi-strategy money management" and didn't get useful result. Sorry if I am asking duplicate questions.

    I am currently using AmiBroker. I think it can try a single strategy on multiple symbols using the same pool of fund, but I don't think it can trade multiple strategies on the same account. The software packages are a little over my budget, but I will keep them in mind.

    I have a feeling that anything involving portfolio re-balancing or forecasting is way too overkill for my personal trading and backtesting. They seem to be more applicable to institutional funds managing a much larger portfolio. But I can do some research and see if any idea can be applied to a personal portfolio.