7. The limits of any screener: what a list of tickers does not tell you

Level 4: Stock Screening

After building three different screeners, it is worth stopping and being honest about what a screener can and cannot do for you.

A screener answers one very specific question: which shares meet, today, these numerical conditions? It does not answer questions such as whether the company is well managed, whether its competitive advantage is sustainable, or whether there is a regulatory or legal risk that shows up in no historical price. A screener can quite happily hand you a list of technically flawless shares belonging to companies with serious underlying problems that the price does not yet reflect.

Nor does it remove survivorship bias if you are not careful: if your starting universe is "the shares that are in the S&P 500 today", you are by definition excluding every share that was thrown out of the index for poor performance — your screener will never see them, because they are no longer in the starting list. The result can look better than an honest backtest over the real universe of shares that existed at each point in the past.

And a screener that combines many conditions — such as Minervini's with its seven simultaneous filters, or the third one with MACD, RSI and P/E all at once — has a silent statistical problem: the more conditions you demand at once, the fewer shares survive, and there is a temptation to keep adding or tweaking conditions until the screener returns exactly the results that "confirm" what you already expected to find. It is the same data snooping trap you will see formalized later in the course, applied here to building filters rather than to backtesting strategies.

None of these limits means a screener is useless. It means its output is the beginning of an investigation, not the end: a short list of candidates that deserve real time from you, not a ready-made shopping list.