Let's say I screen the stocks in the S&P 1500 using the following criteria (all for the last twelve months): Sales greater than enterprise value; sales growth between 0% and 15%; net profit margin greater than 5%. Tested on Portfolio123 since 1/1/2000 with a four-week holding period and a slippage of 0.25% per transaction, this screener gets a pretty nice annualized return of 14.99% with an average portfolio of 48 stocks.
Now let's say I run another screener for S&P 1500 stocks. I want stocks with falling earnings: The current quarter's EPS estimate should be lower than the GAAP EPS in the same quarter last year. I want stocks with a trailing-twelve month unlevered free cash flow of less than zero, just to make sure they're completely worthless by DCF analysis. And I want stocks with decreasing volume: Last week's average volume has to be lower than the average volume three weeks ago. This is a horrible group of stocks. Tested since 1/1/2000 with a four-week holding period and a slippage of 0.25%, this screener gets a terrible annualized return of -1.65% with an average portfolio of 36 stocks.
Now let's take the case of General Motors (NYSE:GM). It passes the first screener with flying colors. It has sales of $170 billion and an EV of only $117 billion, making it a real bargain. The company has a nice healthy profit margin of 5.92%. Its sales this year went up by 10.66% compared to last year.
GM, however, also passes the second screener with flying colors. The EPS estimate for the second quarter is only $1.69, less than its GAAP EPS of $1.81 this time last year. The company's capital expenditure is almost $10 billion more than its cash flow from operations. And its average volume last week (10 million shares per day) is significantly lower than it was three weeks ago (16 million shares per day).
Have you ever wondered why, when you create a stock screener, your results are never nearly as rosy as your backtested results? This is why. Every stock can be chosen by an infinite number of stock screeners that screen for different criteria. Some will show that the stock is a sure-fire winner, others will show that it's a sure-fire loser. There's no way to evaluate a stock - whether it be GM or any other - solely on the basis of a screener. This is what I call the paradox of stock screeners.
The lesson? Don't use stock screeners to pick stocks.
What you want to do is to evaluate every stock you choose from as many angles as possible. Screeners can't do that. If you use more than five or six criteria in your screener, you're only going to get five or six stocks that fit them all. What about all the other criteria that are important to stock market performance? No screener can cover them all.
But there is a way to solve this problem. Instead of binary screening criteria, use ranking systems.
Here's the difference. Let's say you have six equally weighted factors. A screener will accept or reject each stock depending on how you set the limits for each factor. A ranking system, on the other hand, will assign a percentile rank to each stock and then average those ranks to get a final number. The stocks with the highest ranks will have the highest average percentiles of the factors. A stock screener can only look at a few factors; a ranking system can take into account a huge number of them.
Let's look, for example, at a ranking system based on the criteria I used to set up the two screeners above. I'm looking for stocks with a high ratio of sales-to-enterprise value; an annual sales growth that's pretty close to average; a high net profit margin; high earnings growth based on this quarter's estimate compared to the same quarter last year; a high ratio of unlevered free cash flow to enterprise value; and increasing volume. So I'll set up a ranking system that weighs each of these things equally. Here's the result of investing in this ranking system by decile:
And if you just bought the top ten stocks in the S&P 1500 from this ranking system, rebalancing monthly since 1/1/2000 and paying 0.25% slippage, you’d make a nice annualized return of 17.47%. That's not bad. But just imagine how much you could improve your results if you added other factors to your ranking system - and unlike with screeners, there's no reason not to. You can load your ranking system with absolutely everything you consider important in a stock, and it will serve as a de facto stock evaluator. I recently created a ranking system for S&P 1500 stocks whose top ten gets an eighteen-year annualized return of 37%. And if you broaden your universe to include microcaps and nanocaps, ranking systems can be far more profitable than that.
I'm not saying you should never use stock screeners. In fact, I screen out stocks with low liquidity, companies from corrupt countries, certain industries, and firms that are filing late. But if I can rank them on a factor, I don't screen them on that factor.
I am not a wizard at financial accounting. I have not trained to be a stock analyst. I just know that if I'm going to spend tens of thousands of dollars on something, that something better be good. I treat every investment like a precious object, evaluating it from forty different angles before I buy it, and selling it, usually at a profit, only when I need the cash to buy an even more precious object.
Using stock screeners, I not only lagged my benchmark, I lost tens of thousands of dollars of my hard-earned money. But since November 2015, when I suddenly understood that stock screening relied on a fallacy and started using ranking instead, I've made an annualized return of 48%. Throw out your screeners and evaluate your stocks. You won't regret it.
Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article.