THE ETF PROFIT SELECTION SYSTEM
I am pleased to announce a new computerized ETF selection model that my staff and I will use to uncover winning ETFs in the months and years ahead.

Using a computer to help select only the best ETFs to own makes a lot of sense. The computer never feels hope, fear, or greed. It just crunches the numbers and tells you what to do.

After countless hours of programming, testing and re-testing, I have created a highly profitable computerized model that effortlessly tells me which ETFs to buy (and when), which to sell, and whether or not I should allocate funds to cash, fixed-income ETFs or equity-linked ETFs.

RELATIVE STRENGTH MEANS OUTSTANDING PERFORMANCE
The essence of the system is based on the concept of Relative Strength. I covered relative strength in the "Educational Bonus" section of our June 30, 2003 ETF Authority issue. If you're unfamiliar with this concept, then I'd urge you to read that educational piece before continuing any further. You can find it at the following link: http://www.StreetAuthority.com/n.asp?p=s&d=85&t=s#8

The goal of this new system is to identify a select group of ETFs that will outperform the broader market over the long haul, and to boost our returns by changing that basket periodically as market conditions dictate. To accomplish this goal, I've developed a unique computerized model that will provide us with buy signals on ONLY the best-performing ETFs in the market while ignoring the laggards.

The key to developing this system was to determine what attributes make a top-performing fund. For example, we could have chosen to simply hold the fund that delivered the greatest returns in the most recent one-week period. Unfortunately, holding one fund, even if it is an ETF, can be very risky. I would never want to put all of my eggs in one basket. Additionally, this type of simple system would lead to a great deal of weekly trading, which is something that I do not want to do with our longer-term-oriented Model ETF Portfolio. This new model will help us avoid several pitfalls that tend to eat away at trader's profits -- mounting transaction costs and slippage.

I WILL BUY NO FUND BEFORE ITS TIME
The second problem with basing trading decisions solely on the most recent week's returns is that you increase the risk of buying funds just as they've peaked. That is, the fund you end up purchasing may be overbought (and therefore due for a correction). While it might continue to perform admirably for a period of time after reaching the top slot, at that point there is substantial risk for a pullback relative to other funds. After all, no single fund can outperform its peers forever.

My proprietary software is designed to lock on to the best performing funds and to hold them until they fall from grace. If a given ETF weakens for a week or two, then that is not a good enough reason to sell it. However, when it slips substantially relative to the rest of the fund universe, my system will move on to better-performing funds.

THE BASIC RECIPE
How do I manage this feat?

• Only purchase funds that meet a set of strict performance criteria.
• Hold chosen funds until they fall into the bottom 50th percentile (as determined by my performance criteria) of all funds in our investing universe.

The general metrics I use are:

• Before my system will give us a buy signal on any given fund, it must first rank in the top 20% of all ETFs based on my proprietary ranking formula.
• My system will only remove a fund if it falls below the 50th percentile of all funds (again, based on my ranking formula).
• The system will include dividends into all calculations, not just prices.
• It will incorporate multiple time frames when determining percentile rank (not just the latest week).
• It will include many different types of funds for possible addition to our portfolio:
  -- Broad market indices
  -- Different investment styles (growth versus value)
  -- Different asset classes (stocks versus bonds)
  -- Geographic diversification (International stocks)
  -- Sectors (industry-based ETFs)

SOMEBODY ELSE CAN BUY THE FUND DU JOUR
I make no attempt to find the fund du jour. The model does not simply chase the best performer over the past week. It weights total return over several time frames: one week, four weeks, three months and six months. This will tend to prevent many (but not all) whipsaws. (A whipsaw takes place when you buy a fund, then sell it quickly, usually for a loss. This often occurs when a trader buys an overbought fund, then sells it shortly afterward.)

HOW WELL DOES THE SYSTEM WORK?
Very well, if I do say so myself (more on this below). And to make trading easier, I've limited our fund universe to only the most liquid ETFs on the market. My system only considers about 25 funds for inclusion in our portfolio, and it will never provide us with buy signals on more than five funds at any given time. In addition, from time to time the portfolio may include substantial weightings in cash and/or fixed-income securities. This will help us to outperform the broader indices during steep bear market downturns.

After back-testing this unique computerized system with data starting in 1999, I'm pleased to announce that the model has performed admirably. Also, it's worth noting that I have in no way optimized the system's parameters to achieve above-average gains for this historical time period. I chose timeframes and inputs that made sense based on my in-depth knowledge of relative strength principles and used one set of rules to make the fund's selection decisions. Because it is based on a variety of important trading rules that have stood the test of time, our new ETF Profit Selection System should continue to perform as well in the future as it has in our historical testing period.

TESTING THE MODEL WITH REAL DATA
Since the vast majority of ETFs did not even exist before 1999, I did not have enough funds to test our model out effectively until September 1999. Therefore, I began testing our model with historical data starting at that point in time. This time period includes the huge run-up in the Nasdaq until the market topped in March 2000, as well as the entire bear market and its recent recovery. In other words, we saw a wide variety of different market conditions (extreme bull markets, bear markets and everything in between) over that time period. In my mind, that makes it an excellent period in which to test our model.

After running three year's worth of historical data through our model, I'm pleased to inform you that my new ETF Profit Selection System outperformed the broader market every step of the way. Here are the highlights (data through September 19, 2003):

Note: I decided to show you the system's returns since mid-May because that is when I started tracking our previous Model ETF Portfolio. As you can see, despite the range trading that has occurred since June, our new computerized model has easily outpaced the returns posted by the S&P 500. Also, although the model did show losses in 2001 and 2002, its returns were far superior to the overall market.

A $100,000 portfolio placed in this model in the middle of September 1999 would be worth more than $130,000 today. By contrast, the same funds placed in the S&P 500 would be worth less than $90,000! And if you had made the mistake of putting all of your eggs into one basket -- the Nasdaq Composite -- then your portfolio would now be worth less than $70,000!

DOES THE MODEL PROVIDE STOP LOSS LEVELS?
No. This new model stays in all funds until they no longer meet my holding criteria. These criteria are updated once each week following the market close on Friday. All trades are executed at the opening bell on Monday.

HOW OFTEN DOES THE MODEL TRADE?
The average holding period for any given fund is 7.3 weeks, but it has been 6.9 weeks since March of 2002. That time period included several sharp market trend changes that led to increased trading activity (July 2002, August 2002, October 2002, November 2002 and March 2003).

The bottom line here is that our new system should, on average, tend to generate about two or three trades every month. However, depending on market conditions, we might go several weeks without a single change to the portfolio. In fact, as of the close of trading last week, the portfolio had been steady for 10 straight weeks.

ARE ALL ETFS THAT YOU FOLLOW IN YOUR ETF RELATIVE STRENGTH MONITOR ELIGIBLE FOR INCLUSION IN YOUR NEW MODEL PORTFOLIO?
No, I have not included all of the ETFs that I track in our Relative Strength Monitor. To be specific, I decided not to include any of the Merrill Lynch HOLDRS in our new model. There are several reasons for this:

• Several HOLDRS pay dividends every month. This can lead to larger-than-desired cash balances.
• HOLDRS will pay you in stock if one of the companies in the fund completes a spinoff. For example, if Wal-Mart (WMT) decided to create a tracking stock for Sam's Club, then Sam's Club would probably not be added to the Retail HOLDR. Instead, you would receive a distribution of Sam's Club shares. This would complicate our analysis and would lead to undesirable transaction costs.
• The HOLDRS mimic other funds we already include in our model.
  -- The idea of our selection method was to choose from among a diverse array of different funds.
  -- Most of the HOLDRS that trade in significant volume are part of the tech sector and thus tend to move in tandem with each other. This increases the volatility of returns and adds little diversification.
  -- We already incorporate several ETFs that closely track the tech sector (XLK, QQQ, IWM).

Here is a list of all ETFs that I will include in our weekly selection model:

SUMMARY
This week's report heralds the introduction of our new ETF Profit Selection System. It is a computerized model that attempts to choose the best performing exchange-traded funds by employing a proprietary formula that I developed. This formula is based on the concept of Relative Strength.

You can find this week's initial selections in our Model ETF Portfolio section above. Please note that all trades will be "market on open" trades for the first trading day of the week. I will use 9:30 AM as our entry time and will not place orders during pre-market ECN-only activity.

I sincerely hope that you enjoy (and profit from!) this new-and-improved feature of my ETF Authority newsletter in the months ahead.

Good trading!

Steven Poser
Editor
The ETF Authority
New York, NY