Multiple Strategies

P

Paccc

How many people here trade more than one system or strategy at the same time? If you do, do you pick which one you use for each day, and trade only that one for the whole day (such as a reversal system one day and a continuation one another)? Or do you have both systems running concurrently? How do you determine how much capital each system gets to use, does each get half the equity to work with and do position sizings and risk management off its portion, or off the entire portfolio?

Is trading multiple systems better than a single one? I assume that just like diversifying your portfolio with multiple positions is good, so is diversifying your strategy with multiple systems to average out the risk. But if you have a single strategy that performs very well, is it worth working on a new one? Finding a profitable edge is hard enough to do once, let alone multiple times.

For those of you who have experience with this I would be interested to hear what you have to say. I am planning to develop a system, and am wondering if it would be better to design each one to work in tandem with another, or to make one that "does it all." Designing separate ones might allow each to take larger risks on any one position in the hopes that another strategy is playing it more conservatively, etc. Let me know your thoughts on this, and how you chose to design your basket of strategies. Thanks!

-- Paccc
 
Yes, a portfolio of trading systems is a common strategy used to offset performance decay of any one system. The idea is to isolate the best trades though, not to overtrade. An exc book on system dev is "New Trading Systems and Methods", 4th Ed, Perry Kaufman, well worth the money if you're serious.
 
Thanks for the input, I remember seeing that book at the bookstore before and flipping through it. Its a pretty hefty book if I remember correctly. My initial impression of it was not that great, but I shall definitely have another look, thanks!

-- Paccc
 
I'm another one who trades multiple simultaneous strategies. Software backtesting has convinced me that it increases the gain-to-pain ratio significantly. Real money results have been satisfying too.

Unlike a previous poster, I take every signal from every system. This means I occasionally have contrary positions in the same tradeable: system A is long, system B is short, and system C is out. Once you figure out a bookkeeping method that lets you track this easily, it's not especially difficult or confusing at all.

For those who are in love with the idea of non-correlated or anti-correlated systems, let me ignite some controversy by making the following OBVIOUSLY TRUE remark:

If system A only takes long positions, and if system B only takes short positions, and if A and B trade the identical same portfolio, then the returns of A and the returns of B are guaranteed to have a negative correlation coefficient. Voila! Nirvana, served up for free.
 
Quote from horribilicus:


If system A only takes long positions, and if system B only takes short positions, and if A and B trade the identical same portfolio, then the returns of A and the returns of B are guaranteed to have a negative correlation coefficient. Voila! Nirvana, served up for free.
More...

Just checking, you're suggesting this is a good thing to have in practice, rather than being an illusory piece of statistical trickery?
 
Quote from horribilicus:

For those who are in love with the idea of non-correlated or anti-correlated systems, let me ignite some controversy by making the following OBVIOUSLY TRUE remark:

If system A only takes long positions, and if system B only takes short positions, and if A and B trade the identical same portfolio, then the returns of A and the returns of B are guaranteed to have a negative correlation coefficient. Voila! Nirvana, served up for free.
More...

Guaranteed "negative correlation"?

A's long can be B's short constantly, isn't it true?
 
Quote from mahras2:

When people talk about non correlated systems typically it means in terms of returns.
More...

Taking A's +tive signals plus B's -tive signals wouldn't attain what we want (whatever it's called, correctly or not), imo.
 
Quote from horribilicus:

For those who are in love with the idea of non-correlated or anti-correlated systems, let me ignite some controversy by making the following OBVIOUSLY TRUE remark:

If system A only takes long positions, and if system B only takes short positions, and if A and B trade the identical same portfolio, then the returns of A and the returns of B are guaranteed to have a negative correlation coefficient. Voila! Nirvana, served up for free.
More...

No. Although I suspect you knew that.

Obvious example. Both are longish term systems. Choppy market. Spike takes A long ... a day or so later spike down chops A out and takes B short. Then a few days later ... etc until the chop clears.

So you get A and B both contributing to the drawdown.


My approach has been to say:

System A is good and my favourite thing. But when will it lose money. Can I get a System B that makes money overall and will help reduce the pain by making money when A is losing it.
 
The best strategy is to trade multiple system on different baskets of markets. We want systems which are not correlated. One example of this is to trade a trend following system on a commodity basket and then trade a stock index system on the SP500 and Nasdaq mini for example. Since stock index systems are shorter term and often counter trending they will have a low correlation between daily equity changes with the trend following system. A small positive correlation ok because we want both curves to have a upward trend.

Using TradersStudio you can do all of this type of analysis. You can develop systems on baskets of markets as well as single markets and then combine them and develop rules to trade the basket. This type of development of what I call a "Trade Plan" is best way I know to be successful trading long term.
 
You must log in or register to reply here.
Back
Top