EL SEGUNDO, CA (Nov. 18, 1999) -- At one time all of the strategies in the Workshop were the work of its founder, Robert Sheard. That all changed with the appearance of Sparfarkle to the board and his now famous post titled This Strategy Beats UG5. From that post we learned that Value Line data from 1986-1998 was available for a price, and we also learned about the first non-Sheard strategy of the Workshop, the Spark5.
The Spark5 was inspired by James P. O'Shaughnessy's book Invest Like the Best. In Chapter 4 of the book, O'Shaughnessy shows a growth model that is intended to simulate the CGM Capital Development Fund. What Sparfarkle did was make some slight alterations to maximize the return.
Here is the original O'Shaughnessy strategy from page 74 of Invest Like the Best:
Screen the Value Line database for stocks that meet the following criteria:
- Earnings per share (EPS) growth over previous 12 months has been 20% or higher.
- Price growth over previous six months has been 20% or higher.
- EPS growth from last quarter has been 20% or higher.
- Set the value for the estimated EPS growth rate for the coming fiscal year that is low enough that 10 stocks pass the test. The value is somewhere around 30% usually. Vary the number until you get 10 stocks.
Sparfarkle kept the first three criteria the same but limited his screen to the Value Line stocks that were rated 1, 2, or 3 for timeliness. He also set the EPS growth level for the coming fiscal year at 29% or higher and then ranked by size (market capitalization) all of the stocks that passed the test. He tested a strategy of buying the five stocks with the highest market cap (i.e., the five biggest companies) each January and holding for one year.
Obviously, the two strategies are very similar. O'Shaughnessy used the estimated EPS growth variable as a way to limit the list to 10 stocks as the final sorting variable while Sparfarkle kept it as a fixed number. This means that the list generated won't have the same number of stocks each time, but it's much easier to run the screen this way. (Thanks, Sparfarkle!) The most significant difference in the strategies, though, is that Sparfarkle added a final sort by market cap. He did this because he found that the larger companies had better and more consistent returns.
Another difference is that Sparfarkle's strategy is limited to stocks with a timeliness rank of 3 instead of the entire database. How this change came about I'm not sure. However, it may be an unnecessary one because I don't think it changes the outcome. From my research, the Spark strategy will never select a stock with a timeliness rank of 4 or 5 even without the explicit restriction. Any stock with fundamentals that good would have been ranked higher by Value Line, anyway.
How does the Spark do?
Fantastic. For portfolios starting in January, according to the Gritton backtest engine, the five-stock annual strategy had a CAGR of 37.0% for the test period, 1986-1998, with a standard deviation of 29.9%. For the same period, the S&P 500 returned an average of 17.5% with a standard deviation of 12.2%. A 10-stock Spark Strategy also does well with a CAGR of 30.6% and a standard deviation of 29.6%.
Another nice thing about the Spark5 is that it is not as dependent on relative strength. Most of our other workshop screens tend to use the stock's recent price performance as a major factor. The Spark5 uses it as one of five criteria (the four above plus market cap). Because 26-week return isn't the final criteria, this screen tends to generate a more diversified set of stocks than the other strategies that rely heavily on price momentum. For example, the current Spark5 contains a number of big-name stocks -- such as Intel, Wal-Mart, and Home Depot -- that aren't on any of the other strategies.
The ability to broaden my portfolio while getting high annual returns is one reason why the Spark5 remains one of my favorite annual strategies.
Until next time, Fool on!