The Allure and Illusion of Mechanical Investing: Part 2

Thanks for sticking around with me through part 1. Now that we're here, we'll take a look at some tools for developing a mechanical strategy and some of the more subtle points. Fortunately for you, legendary Fool members mghens and keelix are dedicated to mechanical investing, and they post an updated list of many stock screens on a weekly basis at the Fool mechanical investing discussion board. Mghens lists stock screens based on Value Line data and keelix lists stock screens based on AAII's Stock Investor Pro (SIPRO) data. They also have their own websites here and here (keelix's site requires free registration). Both provide updated weekly lists of the stocks generated by the most popular mechanical screens -- developed over the years in the Foolish Workshop -- and later, on the discussion board.

But, you may ask, what good are screens without the ability to verify their usefulness through backtesting? One backtesting resource comes from another legendary Fool member, Zeelotes, who recently backtested many of the Value Line and SIPRO screens over the eight-year period of 1997-2005. For purposes of the backtest, he assumed that one picked the top five stocks from the screen and rebalanced the portfolio on a monthly basis. The no. 1 ranked screen from Zeelotes' study is a SIPRO screen called "P/S I Love You." The P/S stands for "price-to-sales ratio," as in the lower the better. The full screen definition can be found here.

According to Zeelotes' backtest, the P/S I Love You screen returned over 52% per year, but with a relatively high standard deviation of 32.75%. The performance of the S&P 500 index over the same period was only 3.58%, but the index's standard deviation is much lower, around 15%. Thus, the P/S I Love You screen promises outstanding relative returns, but makes you sweat a roller-coaster ride of volatility before you can enjoy your gains. As of Jan. 27th, the top five stocks from this screen were: World Fuel Services (NYSE:INT); Ryerson (NYSE:RYI); Amerisourcebergen (NYSE:ABC); McKesson (NYSE:MCK); and Tenneco Automotive (NYSE:TEN).

Want to perform your own backtest? Keelix's website allows you to backtest SIPRO screens (he calls them "jobs"), but you either need to know a computer programming language called "Radiscript," written by former Fool Phil Radish, or you must limit yourself to using the predefined Radiscript screens that keelix provides. A more user-friendly backtesting site comes from one of the all-time great Fool contributors, Jamie Gritton, who established a free website, http://www.backtest.org/, to backtest Value Line screens. The backtester goes all the way back to 1986, which allows for robust, long-term results. For example, take a look at the 1986-2004 backtest of the Value Line screen "GAR4" (click on the "Run" tab). This screen requires a Value Line Timeliness ranking of 1 or 2 and a low price-to-book ratio. It then picks the five stocks within this universe that have the strongest near-term price appreciation. The portfolio is rebalanced on a monthly basis. The results of the backtest are shown below:

Over the 19-year period backtested, GAR4 beat the S&P 500 in 17 years, or 89% of the time. Furthermore, its compounded average annual return almost quadrupled the return of the S&P 500! As with the P/S I Love You screen, however, the road to riches was a rocky one, with GAR4's standard deviation at twice that of the S&P's over the test period (34% to 17%). As of Jan. 27, the top-five picks of the GAR4 screen were: Veeco Instruments (NASDAQ:VECO), Newport (NASDAQ:NEWP), Ariba (NASDAQ:ARBA), Tidewater (NYSE:TDW), and CTS (NYSE:CTS).

Given these eye-popping returns, why doesn't everyone invest this way?

I can think of four reasons:

1. Time and effort. P/S I Love You and GAR4 are monthly screens, which require an investor to take the time and effort to rebalance the portfolios on a monthly basis. Many people would rather leave the trading to others, such as mutual funds.
2. Trading costs. Monthly trading entails commission costs, slippage costs (bid/ask spreads), and tax costs. The compounded annual returns provided above for P/S I Love You and GAR4 do not include these costs, thus making their outsized returns unrealistic. You can use Jamie Gritton's backtester to account for these costs by clicking on the "trading simulator" tab. The simulator lets you enter inputs for account size (e.g., $20,000), commissions (e.g., $10 per trade), bid-ask spreads (e.g., 0.25%), and tax rates (e.g., 30% short-term, 0% long-term). After inputting these into the simulator, the annual return for GAR4, for example, declines from 44% to 28%, and the percentage of years it beats the S&P 500 declines from 89% to 63%. Still a good record -- it has more than twice the return of the market and outperforms the market a majority of years -- but not as dreamy as before. As Fool member moebruin outlined back in the halcyon days of the Foolish Workshop, trading costs can overwhelm the gains of monthly screens, especially for small account sizes of less than $20,000. Play around with the "initial investment" box in Jamie Gritton's trading simulator. You'll see that the smaller your investment, the more the trading costs eat into the screen's return.
3. Data mining. People fear that a stock screen with a good track record has performed well out of sheer luck, which means that it won't continue to work in the future. For example, a few years back, a researcher did a computer analysis of a United Nations' database and found a very strong correlation between the performance of the S&P 500 index and butter production in Bangladesh. Obviously, no one in their right mind would risk money in the markets on faith that this coincidental correlation would continue into the future. The screen that most closely fits these criteria is GAR4 because it has over 10 years of superior backtested returns and includes criteria that make sense, such as low price-to-book valuation and a 1 or 2 timeliness ranking from Value Line.
4. Emotional fortitude. How many among us are willing to buy stocks on faith alone? Many investors need to hear a "story" before they invest in a stock. They need to understand why a stock is likely to go up, whether it is because of an exciting new product or a charismatic new CEO. Merely seeing a computer spit out a stock symbol isn't emotionally satisfying enough. For such people, the temptation to "modify" the results of the stock screen, by skipping a stock listed or adding a stock not listed, is irresistible. Of course, by modifying the screen, one is invalidating the backtested results. My advice: Don't fix something that ain't broke! If you can't mentally handle buying the stocks generated by the screen on faith, just stay away from mechanical investing altogether, because modifying a screen without a backtest is just gambling.

So far I've talked about monthly screens, which historically have huge compounded annual returns but require effort and only work with $20,000-plus account sizes because of the high trading costs involved. Is there any hope for the smaller, busy investor? The answer is yes. Yearly screens also exist with good track records. Unfortunately, the screens that work superbly on a monthly basis don't seem to work well on a yearly basis. You can see this on keelix's website, where he shows the backtested results for most SIPRO screens over periods ranging from monthly to yearly. Take, for example, the "78 RPM" screen. On a monthly basis, it averages a compounded return of 79%. In contrast, on a yearly basis, it averages only a compounded return of 16%.

This brings me to another problematic wrinkle with yearly screens: The results can vary dramatically depending upon which month of the year you decide to start and annually rebalance your portfolio. The 78 RPM screen demonstrates this problem clearly. If initiation and rebalancing take place at the beginning of February each year, the compounded annual return is a sweet 37%. However, if the rebalancing takes place at the beginning of July, the compounded annual return is a nauseating -6 %! The simple solution is to initiate the 78 RPM system only at the start of February, but you must have some logical rationale for doing this; otherwise, you could be found guilty of data mining. The better solution is probably to avoid the 78 RPM screen altogether and find a yearly screen that has more consistent results, regardless of the start date.

For stock screens tailor-made for yearly rebalancing -- as opposed to monthly -- I rely on James O'Shaughnessy's book, What Works on Wall Street. He has backtested yearly stock screens over a 40-year period, from December 1963 to December 2003. One of the best screens in the book is a yearly screen called "Improved Cornerstone Growth" which, like the P/S I Love You screen, is based on a low price-to-sales ratio and price momentum. The strategy has a compounded annual return of 21.08%, with a standard deviation of 32.55%. This almost doubles the S&P 500's compounded annual return over the same period, which was only 10.61%, with a standard deviation of 16.82%. And since it's a yearly screen, trading costs occur only once per year and are thus negligible.

OK, there you have my summary of the Web resources available for mechanical investing, as well as the types of annual returns to expect and issues to consider when formulating a mechanical stock screen. The question remains whether you should use any of this stuff in your real-life investing. Stay tuned.

Fool financial editor Jim Fink does not own any of the shares cited in this article. The Motley Fool has a disclosure policy.