Ok, you want to design a screen that will outperform the market. So, you decide to select a dozen or so variables that, in your educated opinion as a financial wizard, are solid financial/accounting criteria. Low debt, steady profits, recent earnings surprises and/or upward revisions, very high forward-looking earnings, decreasing debt/equity ratios, increasing market share, and whose price performance is in the top 10% of all stocks. You then find a screener that is capable of searching the database of all stocks and selecting only the top five or ten stocks that meet your criteria.
You decide that you want to select the top five stocks that came up on the screener after Friday's close. You invest equal dollar amounts of each stock, and hold for exactly four weeks. You also want to compare how five representative ETF's do during the same time period. So you select DIA, SPY, MDY, QQQ, and IWM to get a broad representation of all markets, large and small. You also buy equal amounts of each of these as a basis for your comparison. So, at the end of every week, you select five new stocks and you select the same five market stocks for comparison purposes, using Friday's closing price as your buy price and four weeks later as your sell price for both ports.
At the end of the first four weeks and then every week thereafter, you record the percentage increase of both the screener portfolio and the market portfolio. Say you get the following after one year, in which you have tracked 52 separate portfolios of five stocks, each port. of which was held for exactly four weeks.
You add up all the 52 percentages, some of which were negative while others positive. 52 data points each for both portfolios.
You end up with this:
Average percentage increase for screened selections: +4.0%Standard deviation for these: 11.9%
Average percentage increase for the market ports: +1.1%
Standard deviation for these: 4.5%
Question:
Evaluate this process from a statistical point of view and discuss implications for an investor desiring to make a decision to just play the market ETF's(less risky) or to go with the screened portfolio(more risk but higher returns). What inferences can be drawn from the data? Is there a better way to gather the statistics? Is the process sound? Why or why not?
HT |
