This post is about how I define how much money is assigned to each trade also known as money management. More sophisticated traders define money management about how much risk you want to take for given trade. I want to extend that to how one can reduce risk and improve returns of it’s overall portfolio by using advanced money management rules.
In the past I spend a lot of time to find the best entries and exit for a given system. What really made a difference in my trading was to re-focus from being system centric to portfolio centric. In my past post I wrote about the systems I trade or more specifically the systems that make up my portfolio. Of course it makes a lot of sense to combine a momentum with a mean-reversion system and thereby reduce return volatility in your account. But very little is written across the blogosphere about how to manage your money and risk across trading systems.
Often times when reading about portfolio trading simulation it’s being referred as to trading multiple independent systems and assigning capital on a per system basis. I think this is missing an important. By looking at the equity curve of a system and comparing it to other systems you “lose” a lot of invaluable information. What created the equity curve? The number of system specific trades.
So in reality I put all my expected trades as well as current holdings for a given date into a basket. In a second step analyze the expected portfolio on three dimensions in order to come-up with an amount per trade.
- Correlation: Why looking at correlation? This helps me to identify cluster risk ie not having too much money allocated to one area of the market. I’m building a simple correlation matrix on two time frames (mid-term and long-term) and putting them into context to each other.
Let me give you an example, e.g. my rotational momentum system (long) might come up with AAPL and GOOG. The TAA (tactical asset allocation) system might come up with SPY and GLD. Undoubtedly SPY, AAPL and GOOG have a high correlation. To some degree it’s a similar trade, so treating them as 3 independent trades would make me vulnerable (cluster risk). So you want to understand how ALL holdings in your portfolio correlate to each other. Assets with lower relative correlation will have a higher bet size than highly correlated trades.
- Overbought / oversold conditions: I’m using a proprietary intermediate OB/OS indicator that in-itself is already adjusted for volatility. The value for each individual portfolio position is calculated absolute and relative to each other. Hence oversold positions will be assigned somewhat more capital than overbought positions.
- Volatility: That’s what most traders do. More volatile positions will be assigned a smaller stake. For this I’m using a simple de-trended ATR calculation.
I don’t believe in what I can’t backtest. Unfortunately AmiBroker doesn’t have the built-in capabilities to do true cross system / portfolio simulation. I’ve overcome this limitation by building a number of custom specific enhancements. Contact me on email hassler.blog (at) gmail.com in case you have an interest in buying these. So what’s been the impact on my Portfolio? As you might have noticed, I don’t post backtests of the systems I trade, therefore I created a sample system.
- Assets: GLD (Gold), IYR (Real Estates), SPY (S&P500) , TLT (20+ year Treasuries) , EEM (Emerging Markets Stocks)
- Five positions are constantly in trade (no timing / no filter)
- Period: Jan. 2006 until Nov. 2011
- Weekly re-balancing
The position sizing algo outlined is also used for my Portfolio Trader service. In case you are interested in mirroring some of my trades read here for more information. At this point I want to THANK David Varadi from CSS Analytics. My money management methodology got significantly impacted by his wisdom.