Foolish Workshop
Characteristics of a Good Screen, Part 2
One Fool's Opinion

By Morris Chernick (Workshop)

EL SEGUNDO, CA (Oct. 5, 1999) -- How does one judge a stock screening technique? I proposed the first two of five rules last week. Today I'll present the rest of our screen judging rules for your contemplation and comments.

Rule 3: The screen must work with good but diminishing returns for lower-ranked stocks.

Most of our strategies select five stocks, and some have 10-stock variations. But in most cases there are many more stocks in the final list. When a screen creates a list of stocks that are ranked -- i.e., the first stock on the list is higher in some characteristic than the next, which is higher in that characteristic than #3, etc. -- the stocks farther down the list should also perform well compared to the market, but not as well as those ranked higher.

If criteria X is used for the final ranking, stocks that have more of X should do better than those that have less. I don't expect a straight drop in average return as you go down the list, but the first five stocks should outperform stocks six through 10, AND six through 10 should also beat the market, if the screening criteria are valid.

For a screen that employees multiple steps, each step should progressively eliminate the less successful stocks -- again, in a general way. Thus when you get to your base of stocks from which the final few are chosen, that entire base should outperform the market.

For example, the RS-26 week screen has a base of 100 stocks, those that are ranked first for timeliness by Value Line. Timeliness 1 stocks as a group have outperformed the market for 34 years.

When only a few of the stocks selected by a screen have performed well, it is likely that the returns are only due to coincidence, i.e., a few outstanding stocks that just happened to land at the top of the list pulled up the average. This is particularly true when the backtesting database is limited.

Rule 4: There must be at least 10 years of historic returns.

Actually, I don't feel assured that a screen works unless there is a 20-year history. However, getting 20 years of data in some cases is not only difficult, it's impossible. So with 10 years, I am willing to take my chances. However, I still realize I am taking my chances. Yes, that means that except for the Foolish Four, all of these mechanical screens are "chances." This is especially true because we don't have data on how the screens work in a bear market.

Rule 5: There should be peripheral data that supports the contention that the screens should work.

This data can take many different forms, and the more the better. For example, the Foolish Four meets this test because the Beating the S&P 500 strategy uses the same screening criteria (yield and price) but works with a different set of stocks. Also, I have tested similar results using the Value Line database.

Some of the forms the peripheral data can take are:

1. Similar screens that also work well.
2. A backtest using different holding periods, i.e., one month vs. one quarter vs. one year, all of which perform well.
3. A backtest that shows similar results for portfolios that are started/renewed at different times of the year.