Mike Klein updated 06/07/2009

The HG-Type Screen

The "Hidden Gems" (HG)-Type Screen is named for the Motley Fool "Hidden Gems" stock newsletter, and was originally developed by looking at the company fundamentals of some of the best performing newsletter picks and seeing if it would be possible to write a screen that captured the essence of the dynamics of those businesses, on the theory that small or micro-caps hold some promise for substantially market-beating gains. The screen looks for companies that are in the beginning stages of substantial growth as evidenced by accelerating growth in sales, improving product economics as shown by increasing gross margin, positive and increasing operating cash flow, a decent balance sheet and current asset management, strong insider ownership and limited institutional ownership, and reasonably priced. An 8-quarter history of company fundamentals is used to provide this information and much of the screen consists of looking at year-to-year changes. Turnarounds, growing companies, and stable companies with improving "internal" fundamentals (margins) are all good candidates. Most are very small, under $200 million in market capitalization, but occasionally large and well-known companies pass the screen.

Weekly updates of the screen results are posted to the Motley Fool Dynamic Value Investing discussion board with short analyses of new companies appearing on the list. More detailed analysis of some of the companies is occasionally done to identify good potential investments.

The raw weekly screen output (without further commentary) is posted on Sunday or Monday.

The screen has evolved over time and is on its third major version. Each version, based on historical back testing, has substantially improved results over the previous version. Although the changes have not altered the general nature of the screen, the companies passing change substantially from one version to the next. These pages only report the methodology and results of the most recent version which was put into operation on April 25, 2005.

Screen Description and Methodology

The HG-Type Screen in a nutshell and in technical detail.

HG-Type Screen Portfolios

Two general sets of portfolio results are investigated and reported here: Current and Historical.

Current portfolios consist of stocks picked on a weekly basis starting June 01, 2004 and posted to the Screening For Gems discussion board. A stock is added to a Current portfolio the first time it appears in the screen output in a discussion board post. They are therefore free of any potential "data mining" or back testing related issues and are a reliable indicator of the screen's real-time performance. All the Current portfolios are manually maintained.

Historical portfolios are constructed by mechanically running the current version of the screen on historical databases of fundamental data, and using a simple method of selecting stocks: placing a stock in the virtual portfolio once when it is first picked by the screen. All portfolios reflect purely mechanical applications of the screens described here. All buy prices are the closing prices of the first trading day after the screen's database date.

Note that all portfolio results assume: 1) equal dollar amount purchases of each stock; 2) zero cash balance at all times; 3) infinite funds available to invest (i.e. no forced selling due to capital limitations). In other words, portfolio results are the most optimistic possible.

Although a purely mechanical stock selection method generating high returns is certainly an ideal outcome, it is very unlikely to be successful. The final investment decision needs to be made after in-depth analysis. The flavor of analysis that seems most successful in weeding out losers is to determine whether the factors causing the company to pass the HG-Type screen are sustainable, and not a one-shot, cyclical, or fluke occurrence. This is best done by digging into the company's SEC filings, news releases, etc. for in-depth fundamental business analysis.

Always do your own analysis to determine whether to invest in any of the companies mentioned on this web site! Nothing here is a recommendation to buy or sell any securities, and I am not licensed or approved in any way to give that kind of advice. All decisions you make on the basis of any information here are strictly your own responsibility!

Current Portfolios

The Current Portfolios are manually maintained and give the actual performance of the portfolio picks as posted to the Screening for Gems discussion board, as if someone had purchased the all the stocks mentioned in the posts the first time they were mentioned (and were able to buy the full position at the close price of the first trading day after the list was posted). The Current Portofolios are reviewed regularly to identify and adjust for any corporate actions such as mergers and acquisitions, delistings or relistings, ticker changes, bankruptcies, etc. There is no survivorship bias in the Current Portfolios.

The HG-Type Screen has undergone a number of significant changes since it was first described on 06/01/2004. Although some changes have had significant effects on which companies were picked by the screen, the underlying philosophy behind the screen has remained consistent all along. The Current Portfolios therefore are somewhat of a hodge-podge of different types of screen results, but reflect what an investor might have achieved if they had simply bought the stocks mentioned in each week's post. To help analyze if some versions of the screen provide significantly better or worse results than others, this page describes the significant changes that have been made.

Further filtering of the screen picks is also investigated with the Current Portfolio by manually applying a simple Technical Analysis filter as developed by Motley Fool contributor NukeJohn (described here), called NJTA.

CURRENT (SINCE 06/01/2004)
Updated Daily
HG-Type Screen only
HG-Type Screen, filtered by NJTA
Historical Portfolios

The "Historical" portfolios show results had the current version of the screen been run regularly since January 2003 and August 1997 (the oldest database available) respectively. These portfolios are mechanically regenerated at each update. Two flavors for each timeframe are available, one where all the stocks are held until today ("hold forever"), and the other where all stocks are always sold after exactly a 1-year holding period. The latter may seem like anathema to a value investor, but it is based on the observation that many of the stocks picked by the screen that rise significantly in price tend to do so within roughly 6 months of purchase; sometimes less and sometimes more. The 1-year holding period should allow enough time for the business to start playing out its hoped-for growth or turnaround, attract investor attention, and meet the long-term capital gains holding requirement. Results show a very high historical return for the 1-year sell strategy.

A few simple variations on the screen are also provided based on market cap of the company when originally picked. It appears that small but not tiny companies generally do the best.

Please analyze these results carefully. The "Since 2003" results include at least four generally good years for equities (especially 2003), so this historical analysis does not cover a very diverse set of market conditions and results may well be biased significantly positively. The "Since 1997" portfolio, however, encompasses a broader range of economic and market conditions.

Updated Weekly on Saturdays
Updated Monthly on the First of the Month
All Market Caps
Market Cap over $1 billion
Market Cap between $100 million and $1 billion

Historical Databases, Corporate Events, and Backtesting Biases

The Historical Portfolio results are based on about 400 true point-in-time databases and they accurately represent all (or nearly all) ticker changes, mergers, acquisitions, bankruptcies, delistings, etc. Any companies disappearing from the database while held are denoted by a "Sold" date where there otherwise would not be one. Note that even the use of point-in-time databases, while eliminating the great majority of back testing biases, still have a few important ones which I have listed below:

  1. The prices shown for both buy and sell are the closing prices on the date of the data snapshot, which is usually a Friday. Of course the database snapshot is not available until a day or two later, so it is not possible to buy or sell until the following trading day. It is likely that the full set of one-day movements are mostly random and so may largely cancel each other out
  2. Whatever selection bias is inherent in the database itself. Difficult to estimate this bias, but I consider it likely to be very small.
  3. If a stock is dropped from the database (due to a merger, bankruptcy, delisting, etc.), I consider it "sold" as of the date and at the price of the previous data snapshot. This may not reflect the actual final price. Ticker changes don't affect the screen results
  4. There are no frictional costs. Some stocks are very low volume and low liquidity with high potential frictional costs
  5. No dividends are counted, underestimating overall returns

Of these biases, I believe #4 is likely the most important factor as it is systematic (always optimistic) and potentially large for the smallest companies. #5 is also systematic and may partially cancel #4. I believe all the others are likely to be small and/or random enough to be safely ignored to first order.