What is a good Stock Screen ?


Stock screeners' capabilities are endless but if you don't pay attention, you can end up with troubled strategies.

Building successful Stock Screening gives you a step by step approach to screening.

Here are some additional recommendations to keep in mind when building your screen to ensure it is indeed a good one.


Rationale

There should be a rationale behind your screen, especially why it provides superior results.

For instance, there is a modified Dogs of the Dow strategy that eliminates the top ranked stock (the one with highest Dividend Yield) and double up on the 2nd highest ranked. Backtesting showed that the 2nd highest ranked stock outperformed. Although there may be a rationale as to why the highest ranked stock underperform (maybe it is a true dog), there is no real rationale why the 2nd highest ranked outperform.

It is not enough for a strategy to provide superior results, the returns must be backed by a rational explanation. This will minimize the risk of Data-Mining (adapting a strategy so that it fits past market behavior).

Simple Rules Go back to Top

Favor stock screens that have simple rules. If you see a strategy with 2 dozens screening criteria, watch out. Complex screens are likely the results of Data mining.

Quite often, complex rules come from simpler ones: some people torture the data enough to come up with an optimized set of rules that supposedly provide superior performances than the original strategy. The above Dogs Of the Dow modification is an example.

Pay also great care to criteria that sound "too optimized". If you see PE<23, then pass on. Why not PE<23.7 ?
Round numbers are good enough in stock screening so PE<20 would look less fishy.

Backtesting / Historical Performances Go back to Top

  • Check a strategy's performances against both Bull and Bear markets.

  • Outperformance should not come from few exceptional years but be somewhat consistent. “Consistent" does not mean that a strategy must beat the market every year - NO strategy does this - However many strategies beat the market 3 out of every 4 years and almost every time over any 5-10 years period: this is about the best you can get.

  • Check how your strategy fared over various 5 to 10 years periods rather than how it fared over 30-50 years. The latter may show impressive numbers but very few investors have 30-50 years timeframe.

You do not need to perform backtesting yourself ! Many backtesting results or academic studies are available. James O'Shaughnessy's What Works on Wall Street provides ample backtested strategies, especially for Value Investors. The book provides performances over various timeframes and makes it easy to check a strategy's consistency.

Provided you use strategies that are reasonably similar to academic researches or backtested screens, you should be fine.

To illustrate: James O'Shaughnessy's Cornerstone Growth is actually very similar to his Reasonable Runaways strategy. The main difference is that Cornerstone Growth adds positive growth and relaxes the Price/Sales criterion from <1 to <1.5. In fact both screens have a correlation of 0.96 and the performance difference is not statistically meaningful.

The conclusion is that similar strategies will have similar performances. Starting from Reasonable Runaways, you can build a new screen slightly different by for instance relaxing the Price/Sales. However, you would not need to backtest it. Backtesting is mainly required if you considerably modify existing strategies.

Relative Stock Screening Go back to Top

  • Absolute Screening: PE < 20

  • Relative Screening: PE < Industry Average

Favor relative screening rules rather than absolute ones. Absolute screening may work well over certain periods but may very well prove disastrous in extreme market conditions. Relative screening self-adapts to any market condition so is preferred. This is why Ranking is often used in Mechanical Stock Picking: Ranking is obviously relative.

It is OK to have few absolute criteria but a screen made exclusively of absolute criteria is probably not OK.

Number of Stocks in a Screen Go back to Top

One idea of Mechanical Stock Picking is that a Portfolio’s performances are driven by its factors (Capitalization, Growth, Valuation, Relative Strength, ...) rather than by the individual stocks forming the portfolio.

That is, 2 portfolios with similar factors but formed with different stocks will have similar performances over time.
Therefore, a mechanical portfolio should have sufficient number of stocks so as to have reliable factors.

In addition, the higher the number of stocks, the less prone to Data-Mining your strategy is: since each stock is contributing less to the overall portfolio's performance, any outperformance is more robust and unlikely coming from one or two good picks.

Let’s look at some lower/upper limits.


Lower limit: 5 Stocks
This is the number used in some Dogs Of the Dow strategies.

Tips for low number of stocks:
  • Prefer Large Cap
  • Select stocks in different industries
  • Add a filter on Beta (Beta <1)

Upper limit: 50 Stocks
James O’Shaughnessy used 50 stocks in his mechanical portfolios. You would unlikely benefit with more than that.

In the latest 2005 edition, he backtested the portfolios with 10 and 25 stocks. Performances are improved but standard deviations increased. For Small Cap strategies, O'Shaughnessy still recommends 50 stocks: the volatility with 10 or 25 stocks may be unbearable. For Large Cap, 10 or 25 stocks are OK.

In conclusion: Large Cap screens should have 5-25 stocks while Small Cap screens should have 25-50 stocks.

Conclusion for building good Stock Screens Go back to Top

Good stock screens should have consistent outperformance backed by a rational explanation. They shall use simple and preferably relative criteria. The higher the number of stocks, the more robust your screens will be.

Go Back to Top