BULLETIN #51

The Case for Dynamic Portfolio Selection
Introducing the Market Characteristics Matrix (MCM)

By Chuck LeBeau

It seems that system performance is often directly related to market selection. A system that works well in bonds and currencies may fail miserably in sugar and soybeans. The same is true when trading stocks. In our New York Workshop with Dr. Elder I helped the students design a system for Microsoft that failed when we tested it on Proctor and Gamble. I'm sure these examples come as no surprise to veteran system traders. Portfolio selection has always been a problem. Even advocates of "one system for everything" will have to admit that there are huge variations in performance from market to market with even the best of so-called "universal" systems.

Unfortunately the typical solutions to the portfolio selection process are fraught with peril. One solution is to blindly trade everything in a broadly diversified portfolio and hope that the profitable markets make up for the losers. This solution results in large drawdowns and wasted capital tied up trading in non-productive markets. (Show me a system that can make money in live cattle, soybeans, S&Ps, cocoa and British Pounds and I'll become a willing convert to a "one system fits everything" approach.)

Another solution is to drop all the "bad markets" and just trade the good ones. This is simply a form of curve fitting markets to systems. One obvious problem with this approach is that we don't really know which markets will be the good ones in the future. When I was running my CTA business we had one system and we used it to trade everything. We debated whether we should include live cattle in our portfolio because the historical results of our trend following system in the meats were dismal. We finally decide to throw live cattle into the portfolio just so we would be represented in the meats in case something caused them to trend some day. Also, as bad as the live cattle contract was, it was still better than trading live hogs or pork bellies. We started trading with a portfolio of fifteen futures markets and guess what market was the most profitable in our first year of trading? Sure enough, it was live cattle. Also, over the next few years the British Pound went from being one of our best markets to one of our worst.

Since we don't really know which markets will perform best in the future, a static portfolio, no matter how carefully selected, may not be the best solution. Imagine if we could somehow identify and then quantify market characteristics that made a market "good" and then include only markets that are presently exhibiting those desired characteristics in our dynamic portfolio. When British Pounds were exhibiting desirable characteristics they would move into the portfolio. When they started whipsawing back and forth they would be tossed out of the portfolio until they started trending again. We might trade sugar only once every seven or eight years and catch the big winning trends without having to suffer through the day to day losses of having this difficult market in the portfolio every year.

I know this idea may sound like a "pipe dream" but I think it can be done if we put our collective minds to it. Let's brainstorm a bit here and see where it takes us. I think that certain vital market characteristics can be measured and quantified if we simply give it some careful thought. Hopefully this Bulletin will prompt an active discussion on this topic in our FORUM. Your ideas and suggestions are welcome. Up to now I've spent more time on figuring out the problem and the benefits of solving it than I have on finding the solution.

It seems to me that the important characteristics that we want to measure relate directly to volatility and trendiness. The smoothness of the trend is also a characteristic that would be helpful to measure. We might also want to measure volume and open interest because "good" markets have a way of attracting traders.

We must also keep in mind that not all systems are based on trend-following and what may be a "good" market for trend following may be a "bad" market for counter-trend trading strategies. Therefore some objective ratings of trendiness and volatility are what we need so that we can collect information and then use that information as needed to match the market characteristics with the proper systems.

We also must rely on a major assumption. We must assume that the characteristics that we are identifying will persist for a period of time long enough for us to apply a system that will take advantage of those characteristics. It would be no advantage to merely identify markets that we should have been trading in the past. However, if those markets also produced desirable trading characteristics for a period in the future we would be making a major breakthrough in the evolution of systematic trading. It has been my observation over many years of trading that the desirable characteristics we are searching for do persist for a long enough period for us to identify them and then trade them profitably. I envision a situation where we can add a market to our portfolio and keep it in the portfolio for at least a year or two. I also envision markets that are excluded for five or six years and then added back into the portfolio as they produce good trading opportunities..

My first thought on a solution would be to study and measure range and momentum over various time frames. Time must be an important element of anything we measure. Since we may have systems that trade over a variety of time frames it would be pointless to measure only one time frame. We need to decide those time frames that we want to measure. We also need to decide exactly how to measure range and momentum. My first reaction is to measure in units of Average True Range and then convert the Average True Range to dollars to provide additional information about volatility. For example, if the ATRs are too small in terms of dollars the market may not be one that we would want to include in the portfolio. However, if and when the ATRs expand the market would move into our dynamic portfolio.

Here is a sample of the format that I envision. For now I'm going to call it the "Market Characteristics Matrix (MCM)".

Characteristic A: Perhaps trendiness? (Scale of 1 to 10)
Time frame x = 7
Time frame y = 5
Time frame z = 6

Characteristic B: Perhaps range? (Scale of 1 to 10)
Time frame x = 8
Time frame y = 6
Time frame z = 5

Characteristic C: Perhaps volatility? (Scale of 1 to 10)
Time frame x = 4
Time frame y = 3
Time frame z = 5

Characteristic D: Perhaps volume? (Scale of 1 to 10)
Time frame x = 9
Time frame y = 8
Time frame z = 7

Characteristic E: Perhaps smoothness of trend?

And so on.

There are many obvious items that need to be decided. What are the other characteristics that we might want to measure? What are the time frames we want to measure? How do we set our scale for each characteristic? Do we total the results or do we isolate particular items relative to the system we intend to trade?

Once we have finished this "Market Characteristics Matrix (MCM)" we could pick a system and run the system and run the MCM on past data. We would want to calculate the MCM on markets that have performed well and those that have performed poorly to see if we can isolate those characteristics that determine the success of our system. Once we know the characteristics we need we can select the markets that currently display those characteristics and we have the beginning of a dynamic portfolio that can be updated as needed.

I hope that this will be the first of a series of articles on this topic. I encourage our knowledgeable members to start contributing ideas and opinions on this topic via the FORUM. I think we can come up with something that may benefit us all.

* * * * * * * *
April Workshop in Los Angeles

We have scheduled our next Workshop for April 20 and 21 in Los Angeles.

Our February Workshop was a sell out and one of our best ever. We had a great group. I'll be happy to put you in touch with past Workshop attendees if you would like a reference as to the quality and benefits of these inexpensive (only $1500) Workshops.

Learn How to Design, Test, Evaluate and Implement Profitable Trading Systems. Whatever your market and whatever your time frame you are certain to benefit from this carefully prepared two-day workshop. On the weekend of April 20 and 21, Chuck LeBeau will again teach a small group of traders the step by step process of how to design better trading systems.

It doesn't matter if you trade stocks or futures or if you are a day trader or a position trader. It doesn't matter if you use TradeStation, MetaStock, Trading Recipes or Behold. The procedures that need to be followed are the same. You will learn what you need to know to create and improve your systems at this informative Workshop.

Contact Chuck LeBeau at (310) 265- 9776 for more information or go to
http://traderclub.com/workshop.htm Be sure to make your reservation early to guarantee your spot. Payment can be made at a later date.

Can't attend the Workshop? Consider our Workshop Video Package. Two videos, a video workbook, and an ADX based stock trading system for only $250.00. Go to http://traderclub.com/video.htm for details. If you purchase the video and then want to attend the Workshop we will deduct the price of the video from your Workshop fee.


Good luck and good trading.


* * * * * * * *
Visit our FORUM discussion group and share ideas with other traders.

The Traders Club Forum is the best discussion group for traders anywhere on the web. We have many lively and informative threads in progress that will increase your knowledge and generate helpful ideas. If you are new and need questions answered this is the place. You will find that our Club members are knowledgeable and very helpful. It's a very friendly and supportive group and you will be surprised at the amount of information and assistance that is available. The discussions are very high quality and best of all - IT'S ALL FREE.

Just go to http://www.traderclub.com/discus/board.html and take a look.