Our Simple Starter Screen
Figure 1 refreshes our memory of the screen we used in Chapter 1 to introduce the topic of screening.
Although it backtested well, we identified one serious flaw, the role penny-stocks played in boosting its performance record. Figure 2 shows a modified version that addresses this two ways, with a minimum market capitalization requirement ($100 million) and a starting universe that excludes OTC stocks.
This remains a very simple screen, hardly one that requires the best screening tool one can find. Even so, a backtest shows it edged out the Russell 2000 during the challenging period we experienced over the past year.
Figure 4 shows that it also beat the Russell 2000 over a longer backtest, one that started 3/31/01.
Our ability to achieve successful backtest results based on tests consistent with common sense (the notion that good sales and EPS growth rates lead to better share price performance) and with such a naive screen would seem to reinforce the notion that screening is all we need.
Not so fast . . .
OK. We have a good screen. Lets buy the stocks, all 400-plus of them!
Not every screen produces lists this size. Some have bigger result sets, but most lists are smaller. But they are rarely just right. As I screened over the years, I usually aimed for lists with about 50 stocks. Since this can never be precise, targeting smaller lists can be risky, leading to inadequate diversification or even times when a screen produces no stocks at all.
Because of lists that were usually too big to be bought in their entirety, during my tenure at Reuters.com, I wound up devoting far more elbow grease to figuring out ways to choose from a screening result set than I did creating screens.
A practical solution
Let's see if we can find a way to make our starter growth screen a bit more usable.
Figure 5 shows an area of the screening interface we briefly introduced but haven't focused on, until now; the rank area. It allows us to mix ranking systems with screens. For now, we'll focus on the simple approach, the "Quick Rank." This is where you can articulate any factor or formula that can be used as a basis for sorting a list of stocks (make sure you remember to use the dropdown menu to tell the application whether higher tallies are better, a descending sort, or lower number are to be preferred, an ascending sort).
I used a factor that represents the one-month change in short interest. Whether one is a contrarian who prefers higher tallies that indicate increasing bearishness or one who wishes to go with the flow by preferring lower numbers that result from short covering is a matter of philosophy. I chose the latter; I wanted to sort the growth-screen stocks on the basis of short covering.
This does not change the screen. I'll still see the same 400-plus stocks on the list.
But specifying a quick rank allows me to refine the backtest. Figure 6 shows how I do that.
This is the right side of the screening interface. The second choice, "No. Top Stocks," is set to zero by default. That means the backtest will assume all 400-plus stocks on the lists will be bought (with rebalancing done every four weeks), an impractical approach for most investors.
We see above that I wound up changing the default selection from zero to 25. Now, the backtester will select only 25 stocks from those that passed the screen. Which 25? The top (best) 25. And we define top based on our quick rank. In this case, each rebalancing will select the 25 stocks having the greatest declines in one-month short interest.
Now, we have something that is much more practical; list sizes that better reflect what an investor can own.
Figures 7 and 8 show the results of a one-year and long-term backtests respectively.
In both cases, performance looks better than what we saw for the 400-plus stock lists. One who wanted to use this as a basis for investing could simply create a screen report, sort it based on the one-month change in short interest, and buy the 25 stocks having the biggest declines.
The benefits of the quick rank
There are countless way one can sort a list. Above, we saw just one of them.
Because the Portfolio123 Quick Rank feature allows us to integrate the result of a sort into our back tests, we can and should search for sort criterion that enhance our screening efforts, as we've done here when we sought not just growth stocks, but growth stocks that had experienced recent short covering.
Hence this simple Quick Rank/sort helped us in two ways.
- It produced backtest results that show the potential for better performance than we could get from the screen alone
- It produced a manageable list size based on our choice. Moreover, assuming the screen itself backtests well, the reduction in list size would be worthwhile even if performance were unchanged, and arguably, even if performance was modestly lower (a good list one could buy is preferable to a great list that is not really buyable).
But what about ranking?
So far, despite the label Quick Rank, we talked only about screening and sorting. Where do ranking systems come in?
It's here. We already did it!
When all is said and done, a ranking system, no matter how complex and no matter how much mathematical and statistical jargon surrounds it, is a sort criterion; no more and no less. Everything else done in this area relates to the method used to identify the best-to-worst scale.
In this Chapter, our method was quite simple. It was based entirely on a single factor that was conveniently built into the Portfolio123 data-set.
Other ranking systems, including the widely visible commercial ones such as Money Central Stock Scouter, Morningstar, S&P, Value Line, TheStreet.com ratings, Motley Fools CAPS ratings, and so on, are more complex, meaning they have more factors and usually that all or some of their factors are formulas rather than a single data item such as we used here. And sometimes, factors are based not just on data but human work product.
But no matter how much goes into a ranking system, at the end of the day, all the inputs get sliced, diced, mixed, mashed and ultimately combined into a single number applicable to each stock. And once we have that set of numbers, we can use it exactly the way we used our Quick Rank, as a basis to sort a list. Here, the list we used was pre-selected using our growth screen. More often, the list will simply be an entire stock universe.
In other words, if I had enough faith in the power of the short-covering data item, I could used it to sort the entire universe and taken the best 25. (For what it's worth, even this backtested reasonably, but when I took out penny stocks, it didn't fare so well under the one-year test.)
As long as you remember what a ranking system is, a basis for sorting a list, you'll have no trouble keeping your bearings as we move forward to more sophisticated approaches.
In the next Chapter, I'll introduce the structure of a more comprehensive rank which uses more than one factor. Such multi-factor ranks are the type created by most Portfolio123 users and are representative of the kinds of ranking systems created elsewhere. Once this has been explained, we'll consider other advantages of ranking systems, relative to screens, in addition to the two mentioned above.
In the chapter after that, I'll create a simple multi-factor rank. At that time, you'll learn how to use the Portfolio123 ranking interface. In subsequent chapters, we'll see how more advanced ranking systems are created and used.