Foolish Workshop
By
Finally, some serious backtested returns!
Fools have been testing away using these tools, uncovering a variety of screen permutations with various numbers of stocks, holding periods, and starting times. With this wealth of information, it has become quite a challenge to keep up. In the interest of making life a bit easier for all you Workshop Fools out there, today we are presenting the results of backtesting four of the screens that we follow regularly. The same information will be added soon to the Screen Explanations page. The screens tested here are those that the database supports.
Jamie Gritton's backtester includes the screening parameters used to select stocks each month for each of the screens listed and can provide returns for those stocks at 1-, 3-, 6-, and 12-month intervals, 1986 through 1998. (We don't use the longer holding periods.) It is a monthly database, so we can run each screen as though a new portfolio were started each month of the year.
(To ensure accurate comparisons among holding periods, each screen covers the same total number of months. Because we do not yet have the returns data from 1999, every strategy finished at some time in 1998. For example: A strategy starting in May of 1986 would end in April of 1998 and cover a full 12 years. Each strategy's end point is therefore exactly 12 years after its starting point. Strategies starting in January end in December of 1997. They do not renew in January of 1998 because that would give them more than 12 years. This gives us an apples-to-apples comparison among screens.)
For each strategy, the CAGR was calculated for 5- and 10-stock versions renewing every month, quarter, half year, and year.
Each return quoted below is the average for 12 portfolios following that strategy, i.e., the Beta Quarterly Strategy is the average for 12 portfolios, the first starting in January 1986 and renewing in April 1986, June 1986, etc., every three months through December of 1997; the next starting in February 1986 and renewing in May 1986, July 1986, etc. through January 1998, etc. through the final portfolio that started in December 1986, and renewed every three months ending in November 1998.
We have data available to display the returns for each starting month separately, but I believe that most of the fluctuations between starting months are due to chance and we might as well ignore them. Several screens appear to have better returns when started in January, however, so I've chosen to show the January returns for annual strategies separately. (The regular annual returns include the January numbers in their average.) But I wouldn't trust those January numbers too much. It's better to lower your expectations.
OK, it's time to let the screens speak for themselves.
Beta Screen: Taking the Timeliness 1 stocks in the Value Line Investment Survey, simply sort these stocks in descending order by their beta. Select the 10 (or 5) stocks with the highest betas. Beta is a measure of a stock's price volatility.
Ann.
# Monthly Quarterly Semiannual Annual (Jan. start)
5 18.8% 17.8% 19.3% 18.0% 26.3%
10 21.7% 20.4% 19.6% 16.8% 21.5%
The Beta strategy has mediocre historical returns. It shows no significant preference for any particular holding period, although the January annual screen is interesting. In several instances the 10-stock variation seems to perform better than the 5-stock version. As Moe Chernick suggested in his article about the characteristics of a good screen, "The screen must work with good but diminishing returns for lower-ranked stocks." The Beta screen doesn't have this quality, and I don't think it is a particularly good screen.
Ann.
# Monthly Quarterly Semiannual Annual (Jan. start)
5 34.9% 35.2% 35.7% 30.6% 46.9%
10 27.2% 26.9% 28.8% 26.7% 41.8%
I've chosen to present both the 5- and 10-stock variations for PEG although Moe Chernick, its creator, never intended it as a 10-stock screen. Yet this illustrates the screen's behavior with respect to the aforementioned characteristic of a good screen -- pretty nice, and it really is not bad as a 10-stock screen. What do you think, Moe?
Ann.
# Monthly Quarterly Semiannual Annual (Jan. start)
5 34.8% 32.9% 31% 25.8% 27.5%
10 30.7% 29.6% 28% 24.4% 31.7%
The RS-26 returns seem to be better with shorter holding periods. But before jumping into the monthly screen, read Moe's excellent article about the true costs of monthly screens.
Ann.
# Monthly Quarterly Semiannual Annual (Jan. start)
5 24.7% 28.4% 27.3% 25.1% 32.5%
10 22.1% 23.7% 24.7% 21.7% 27.4%
Sparfarkle, like Moe, never intended this as a 10-stock screen, but the numbers are instructive. We see the expected decline in returns when we go from 5 stocks to 10 stocks. We see no advantage to a short holding period. To the contrary, no one would likely select Spark5 as a monthly screen. And we do see quite an advantage to a January start.