How to choose a stock screen template
Choose a stock screen template by starting with the business question you want to test. Use Piotroski F-Score for improving financial condition, Greenblatt Magic Formula for relative cheapness plus return on capital, Graham Net-Net for balance-sheet discounts, Buffett Quality Value for cash-generative quality at a reasonable price, and Earnings Yield + Quality for absolute operating yields with cash conversion.
Those aren't competing routes to a universal list of good stocks. Each template defines value differently and asks different evidence to clear its gate. Open the one whose claim matches yours, inspect its universe and missing-data rules, then save your own version before changing anything. A template is a worked argument. Treat it as one.
Which question does each template ask?
Piotroski F-Score asks whether a company's filed numbers are improving across several dimensions. Quantery's bundled version groups binary signals into profitability, earnings quality, balance-sheet health, and efficiency with dilution. It uses a demanding score gate. This is the natural starting point when your hunch sounds like, “The business is getting healthier.” It isn't a valuation thesis by itself. A high score can describe an expensive company just as easily as a cheap one.
Greenblatt Magic Formula asks which companies combine high operating earnings yield with high return on capital relative to the rest of the eligible universe. Its core is a combined percentile rank, so the answer is comparative. A company can qualify because it is among the better available combinations even when the market as a whole looks weak. Start here when your claim is, “I want the better balance of cheap and productive within this opportunity set.”
Graham Net-Net asks whether the market values the equity below a conservative balance-sheet floor. Net current asset value is current assets minus all liabilities. The template then checks short-term solvency, cash burn, and dilution because a discount to a shrinking asset base isn't much of a floor. This is a deep-value question about realizable assets. It isn't a general low-multiple screen.
Buffett Quality Value asks whether a business earns well on shareholder capital, generates positive free cash flow at an acceptable yield, carries manageable leverage, and sustains revenue. Use it when the claim starts with durable business quality and then asks what price the equity commands. Its quality measures don't travel cleanly to every sector, which is why the template excludes financial companies and real estate.
Earnings Yield + Quality asks whether operating earnings are cheap against enterprise value, productive against capital employed, and backed by cash. Unlike Magic Formula's relative ranks, it uses absolute bands. That makes it useful when your claim names an economic hurdle: the operating yield and return on capital must clear levels you chose, even if few companies do.
The labels overlap. Magic Formula and Earnings Yield + Quality both pair cheapness with capital productivity. Buffett Quality Value also includes quality and valuation. The formulas settle the difference. Don't choose by the name you like. Choose by the question encoded in the features.
Relative ranks and absolute thresholds behave differently
The biggest fork is between relative and absolute selection.
Greenblatt Magic Formula computes percentile ranks for earnings yield and return on capital, combines them, and admits the top slice that also clears its earnings-power and trading-liquidity checks. A percentile rank says where a company stands among eligible peers on that scan date. Change the universe and the rank can change even when the company's filing doesn't.
Earnings Yield + Quality uses fixed parameter bands for EBITDA over enterprise value and EBITDA over capital employed. Buffett Quality Value does the same for return on equity and free-cash-flow yield. An absolute threshold says the company must clear your declared hurdle. The scan can therefore return a sparse result when the market offers little that meets it. That's useful information.
Piotroski sits elsewhere. Its signals are directional binaries: positive cash flow, improving liquidity, lower leverage, controlled dilution. It asks whether conditions improved without demanding impressive levels. Graham Net-Net uses a hard relationship between market capitalization and the balance-sheet floor, then surrounds it with protective tests.
Write one sentence before opening a template:
I expect companies with improving filed fundamentals to behave differently after they clear a demanding composite gate.
That sentence points to Piotroski. Replace “improving filed fundamentals” with “top-ranked operating yield and return on capital,” and Magic Formula fits. If you can't write the sentence, you aren't choosing a thesis yet. You're browsing labels.
For a fuller treatment of the distinction, percentile ranks and absolute thresholds behave differently. The choice changes whether your rule always looks for the best available names or sometimes says that none clear the bar.
The universe can decide whether the template fits
Read the universe block before the score. Every bundled template starts with the major US exchanges, a market-cap floor, and sector exclusions. Those settings are part of the economic claim.
Graham Net-Net keeps a lower size floor because balance-sheet bargains tend to live among small companies. Its cash-burn and dilution checks exist partly because that neighborhood is rough. Buffett Quality Value starts much higher and asks for established cash-generative businesses. Greenblatt Magic Formula stays broader because its cross-sectional ranks need an eligible comparison set, while Earnings Yield + Quality starts with larger companies and applies fixed operating hurdles.
Sector exclusions matter even more. The templates leave out Financial Services and Real Estate when ratios such as current assets, free cash flow, enterprise value, or ordinary operating leverage stop meaning what the thesis assumes. Templates built around enterprise value and return on capital can also exclude Utilities. A bank's debt is raw material. A regulated utility's capital base is governed by economics that an ordinary industrial-company return rule can misread.
Don't remove an exclusion just to recover a company you expected to see. First state why the ratio should mean the same thing in that sector. Better yet, write a sector-specific thesis with measures suited to it.
A named index or custom ticker list narrows the question again. As the guide to choosing a screening universe explains, ranks inherit the guest list and missing coverage creates an effective universe inside the declared one. Copying a template while ignoring its universe is changing the experiment before it starts.
Read the data policy before trusting a score
Every template declares required fields and a staleness bound. A required input that is missing prevents the company from qualifying. That's preferable to letting unknown data behave like favorable data, but it also means the result reflects local coverage.
Piotroski makes the issue easy to see. Quantery's version uses eight of the original framework's signals because historical gross-margin coverage isn't dependable across the local lake. It also needs comparable prior periods for its change signals. A filing gap can break or shift a comparison, and a missing signal doesn't earn a point. The eight-signal Piotroski walkthrough explains that adaptation in detail.
Greenblatt Magic Formula uses visible proxies too. Its operating earnings feature approximates EBIT from trailing EBITDA and capital spending. Its capital-employed calculation reconstructs an available-data version from balance-sheet fields. Those choices are written in the template. They aren't footnotes you discover after a good backtest.
Graham Net-Net is strict about same-filing balance-sheet inputs because combining current assets from one period with liabilities from another would create a liquidation floor no filing ever showed. Buffett Quality Value guards negative EBITDA in its leverage ratio so a loss-making indebted business can't look conservatively financed through a negative quotient. Earnings Yield + Quality refuses an enterprise-value yield when the denominator or operating earnings don't support the calculation.
Review coverage information for representative companies before a long historical test, especially when your chosen template needs prior periods or cash-flow lines. The backtest coverage preflight shows what to inspect. A template can be economically suitable and still be a poor test on a local dataset that can't evaluate it consistently.
How should you compare two promising templates?
Sometimes two templates express plausible versions of your hunch. Don't run both and keep whichever curve looks better. That turns template choice into an unrecorded optimization.
Instead, compare the claims before the returns:
- Write the business sentence for each template.
- List the feature that carries cheapness and the feature that carries quality.
- Mark whether each rule is relative, absolute, or directional.
- Check the universe, required fields, and null behavior.
- Name what evidence would make each thesis fail.
Suppose you're choosing between Greenblatt Magic Formula and Earnings Yield + Quality. The first says a high combined rank within today's eligible market matters. The second says specified operating yield, return on capital, cash conversion, and liquidity hurdles all matter. One can keep producing a top group as market levels shift. The other can reject most of the market. That's a difference in the claim, not a cosmetic difference in presentation.
Now backtest both over the same defensible dates with the same benchmark where their universes make that comparison fair. Every return is hypothetical and excludes costs. Read coverage first, then membership and event patterns, then performance. If one wins, ask whether its different universe, relative ranking, or proxy made the difference. A backtest result needs perturbation before confidence, and comparing two preexisting templates is only the beginning of that work.
Make the template yours one change at a time
Once you've chosen a starting point, preserve the bundled version and work from your own copy. Change one parameter or one feature at a time. Give the change an economic reason before you run it.
If Piotroski's score gate feels too forgiving, test a stricter gate and inspect which signal patterns disappear. If Graham Net-Net's burn allowance doesn't match your view of a melting asset base, move that parameter and examine the boundary cases. If Buffett Quality Value gives return on equity too much authority, add return on assets so leverage has less room to dress up the result. Each edit should sharpen the sentence you wrote at the start.
Don't bolt the attractive pieces of every template into one giant score. Return on equity, return on capital, cash conversion, and Piotroski profitability can cast several votes for related facts. Duplicate signals can make a thesis look diversified while one accounting outcome does most of the work.
The bundled templates are good starting points because their assumptions sit in the open. Pick the question that matches yours. Read its universe and data policy. Then change the rule you actually disagree with and rerun it. The best template is the one you can explain before the scan and challenge after it.
Research tooling, not investment advice. Nothing here is a recommendation to buy, sell, or hold any security. Screens, scores, and backtests are informational only; backtested results are hypothetical, exclude costs such as commissions and slippage, and do not guarantee future results. Verify against primary filings and make your own decisions.
Want to try this on your own rules? Quantery is free for 14 days: the full app, no card required.
← All articles