How the Greenblatt Magic Formula Combines Value and Quality

Joel Greenblatt published The Little Book That Beats the Market in 2005, and the pitch was almost suspiciously simple. Rank every stock on cheapness, rank it again on quality, add the two ranks together, and buy the companies with the best combined scores. He called it the Magic Formula, and the book's backtest from 1988 to 2004 showed returns above 30 percent a year.

A number like that earns attention and doubt in roughly equal amounts. Could two ratios really beat the market by that much? Two decades of live results later, we have a fair answer. The ranking logic holds up, the 30 percent a year does not, and the thing that actually decides your outcome has less to do with the math than with your willingness to keep running the system when it looks dumb.

The two numbers

Both metrics are built on EBIT, earnings before interest and taxes, which is roughly the operating income line on the income statement.

Earnings yield is EBIT divided by enterprise value, where enterprise value is market cap plus debt minus cash. Greenblatt prefers this over the familiar P/E ratio because it neutralizes capital structure. A company loaded with debt can look cheap on P/E while being expensive once you count the debt a buyer would effectively assume. EBIT over EV compares operating earnings to the full price of the business, so a levered company and a debt-free one can be compared honestly.

Return on capital is EBIT divided by tangible capital employed, meaning net working capital plus net fixed assets. This asks how much operating profit the business generates from the capital actually tied up in running it. The definition deliberately leaves out goodwill, so a company that overpaid for an acquisition a decade ago isn't punished forever. You're measuring how the underlying operations perform on the capital they actually use, and a business that earns a lot on little capital usually has some competitive advantage doing the work.

The ranking mechanics are simple. Every company in the universe gets a rank on each metric. If a stock is the 5th cheapest by earnings yield and the 15th best by return on capital, its combined score is 20. A stock ranked 100th on cheapness and 3rd on quality scores 103. Sort ascending and buy from the top of the list. Absolute values never enter the decision; the formula only cares whether a company is cheaper and better than the alternatives on the same list.

A worked example

Say a company reports 200 million dollars of EBIT. Its market cap is 1.5 billion, it carries 600 million of debt, and it holds 100 million in cash. Enterprise value is 1.5 billion plus 600 million minus 100 million, which is 2 billion. Earnings yield is 200 divided by 2,000, so 10 percent.

Now the capital side. Say net working capital comes to 300 million and net fixed assets, mostly property and equipment after depreciation, are 500 million. Tangible capital employed is 800 million, and return on capital is 200 divided by 800, or 25 percent. A business earning 25 percent on its capital that you can buy at a 10 percent earnings yield is the profile the formula hunts for, though whether it makes the final list depends entirely on what else is ranked that day.

Every input comes from the latest 10-K or 10-Q, free on EDGAR. EBIT sits on the income statement, and the balance sheet gives you debt, cash, the working capital items, and net property, plant, and equipment. Greenblatt also runs a free screener at magicformulainvesting.com that does the ranking for you, though I'd still spot-check a few names in the filings before buying anything, for reasons covered below.

Why the combination works

Screening on either metric alone tends to backfire. Pure cheapness collects businesses that deserve their price, dying industries and broken balance sheets and earnings that are about to roll over. Pure quality has the opposite problem, because everyone else can see the same wonderful business and the multiple usually reflects it.

The overlap is where the formula hunts. A business with a high return on capital usually has something protecting it, and when that kind of business also shows up cheap, the cause is often temporary. A missed quarter, a sector nobody wants this year, a broad selloff that dragged it down with everything else. Headwinds like that tend to pass, prices tend to drift back toward what the business earns, and the formula is positioned there when it happens. It won't be right on every name, which is exactly why it holds 20 to 30 of them at a time.

What actually happened after 2005

Live performance since the book came out has been more ordinary than the backtest. Depending on the window you measure, Magic Formula portfolios have looked anywhere from roughly market-matching to modestly ahead. Nobody running it live has come close to 30 percent a year.

There are a few honest reasons for the gap. The backtest window covered stretches, like the early 1990s recession and the years right after the dot-com bust, that were unusually kind to value strategies. Publication invited more capital to chase the same stocks, which erodes any edge. And the 2010s were a long, growth-led market in which almost everything value-flavored lagged, formula or not.

None of that means the idea is dead. Independent tests in markets outside the US have generally found that the ranking still separates winners from losers, though the size of the edge varies by market and by study. The sober read is that this is a value strategy with a quality tilt, and it behaves like one. It compounds patiently, and it goes through stretches where it trails the index badly enough to test you.

How Greenblatt says to run it

The book's implementation advice is specific, and the details do real work.

• Hold 20 to 30 stocks. Enough that one blowup can't sink the year, few enough to manage by hand.
• Buy in tranches. Add a handful of positions each month over your first year instead of buying the whole list at once. It staggers your entry points and turns maintenance into a small monthly habit.
• Hold each name for one year, then replace it with whatever ranks highest at that point.
• Mind the tax calendar. In a taxable US account, Greenblatt suggests selling losers just before the one-year mark to book short-term losses and selling winners just after it for long-term capital gains treatment.

The screening universe matters too. Greenblatt excludes financials and utilities because their balance sheets make return on capital close to meaningless (a bank's capital employed doesn't mean what a manufacturer's does). His screener also leaves out foreign ADRs and applies a minimum market cap; the book's tests used a floor of 50 million dollars, and a higher floor gets you a more liquid list.

Data hygiene matters more than people expect. As-reported EBIT can be distorted by one-time charges, restructuring costs, and impairments, and write-downs are especially sneaky on the quality side. An impairment shrinks the asset base, which can make return on capital look spectacular for exactly the wrong reason. Before buying a name off the screen, open the filing and skim the footnotes for anything one-time. Ten minutes per stock catches most of the obvious distortions.

Where people actually lose

Greenblatt's own firm ran a version of this experiment on real clients. It offered professionally managed accounts that bought the list mechanically, and self-managed accounts where clients chose which formula stocks to own. The mechanical accounts did meaningfully better. Left to choose, clients skipped the names with scary headlines, passed on the boring industrials, and leaned toward companies they recognized. In doing so they stripped out the most uncomfortable stocks, and the uncomfortable stocks were carrying the returns.

The second failure mode is quitting. The formula trails the index in plenty of individual years, and in growth-led markets it can lag for two or three years running. Greenblatt argues the strategy keeps working precisely because those stretches shake most people out; if it won every single year, everyone would run it and the edge would get arbitraged away. Knowing that in advance doesn't make year three of underperformance feel any better, which is why so many people bail somewhere near the point of maximum discouragement.

The third is tinkering, meaning overriding the ranks one stock at a time on gut feel. Every override reintroduces the biases the system exists to remove. If you trust your judgment that much, run a judgment-driven process honestly instead of a screen you quietly veto.

Modifications that make sense

There's a difference between vetoing individual stocks and rewriting the rules upfront, then following the new rules mechanically. The second approach keeps the discipline intact, and a few variants are defensible.

• A momentum screen. Some practitioners skip names in an outright price collapse, on the logic that the market sometimes knows something the trailing financials don't show yet. You give up some cheapness in exchange for fewer falling knives.
• Extra quality checks. Stable gross margins and consistent free cash flow conversion help confirm that EBIT reflects real economics and that the cash actually shows up.
• Pairing with the Piotroski F-Score. Piotroski's 2000 paper scores companies on nine simple accounting checks for improving fundamentals. Running the F-Score over formula names splits the list into cheap-and-improving versus cheap-and-deteriorating, a distinction the formula can't see on its own.
• Going international. The same two ratios can be ranked in any market with decent accounting data, which widens the opportunity set and spreads you across markets that fall in and out of favor at different times.

Should you actually use it

The formula fits investors who want a rules-based process without building a quant stack. The inputs are free, the math is arithmetic, and the maintenance is roughly an hour a month plus an annual rebalance. For a part-time investor who wants systematic exposure to cheap, high-quality stocks, it's hard to find anything lower-lift that still counts as serious.

It's a bad fit if you need to beat the market every year, can't sit through multi-year lags, or know deep down that you'll second-guess the list. The edge comes from mechanical discipline, and a screen you override at will is just stock-picking with extra steps.

If you're considering it, run a cheap test before committing money. Paper-trade the screen for a few months, look at the actual names it surfaces, and ask whether you would honestly have bought the three ugliest. If the answer is no, you've learned something useful before it cost you anything. If the answer is yes, commit for five years or don't bother, since anything shorter mostly measures your luck with the value cycle rather than the strategy itself.