TL;DR: The Magic Formula, from Joel Greenblatt's The Little Book That Beats the Market, ranks stocks by combining two rankings: earnings yield (cheapness) and return on capital (quality). Greenblatt's published procedure holds the top-ranked names as a portfolio of roughly 20 to 30 stocks, each for about a year, then re-ranks and repeats. The formula is trivially simple on purpose. The hard part was never the math; it is holding a mechanical portfolio through the stretches, sometimes years long, when it trails the market.
What is the Magic Formula?
Joel Greenblatt, a hedge fund manager and Columbia Business School professor, published the formula in The Little Book That Beats the Market (2005, updated 2010) with an unusual promise: explain a real investing strategy so plainly that his children could follow it. The strategy rests on two claims about what makes a stock purchase work over time:
- Good companies earn high returns on the capital they employ. A business that turns each dollar of plant, equipment, and working capital into a lot of operating profit is doing something valuable, and probably defensible.
- Cheap stocks outperform dear ones on average. Paying a low price relative to current earnings leaves room for error and for re-rating.
Either idea alone is a known style: quality investing and value investing. The formula's contribution is combining them mechanically, so a stock must pass both tests at once, and removing the investor's discretion from the selection step entirely.
How does the ranking work?
Two ratios, computed for every stock in the universe (Greenblatt's published version excludes financial and utility companies, whose balance sheets make the ratios hard to compare):
- Earnings yield = EBIT / enterprise value. EBIT is earnings before interest and taxes; enterprise value is market capitalization plus net debt. This asks: for the price of buying the whole business, debts included, what operating earnings do I get back each year?
- Return on capital = EBIT / (net working capital + net fixed assets). This asks: for the capital the business actually needs to run, how much operating profit does it produce?
Each stock gets a rank on each ratio. Add the two ranks together. The lowest combined rank wins: a stock ranked 10th on earnings yield and 30th on return on capital (combined 40) beats one ranked 1st on cheapness and 200th on quality (combined 201). That addition is the whole trick. It refuses both the wonderful company at any price and the statistically cheap company whose business earns nothing on its capital.
The published recipe then holds a portfolio of roughly 20 to 30 top-ranked stocks, bought in batches through the year, each position held for about a year and then replaced by the new top ranks. It is a portfolio strategy, not a stock-picking shortcut: Greenblatt's own presentation is explicit that individual picks from the list will include losers, and the edge, if it shows up, shows up across the basket and across years.
Why is it built this way?
Three design choices deserve attention, because each solves a real problem:
- EBIT and enterprise value, not price-to-earnings. P/E ratios are distorted by how much debt a company carries and by its tax situation. EBIT over enterprise value compares the operating engine with the full price of owning it, debt included, so a leveraged company cannot look cheap merely by being leveraged.
- Return on capital over return on equity. Equity returns can be inflated by borrowing. Return on the tangible capital the business employs cannot be flattered the same way, which keeps the quality test honest.
- Ranks, not raw scores. Ranking compresses outliers. One absurd ratio, from a one-off gain or a data quirk, can dominate a scoring system; it can only ever be first in a ranking system. Mechanical rules that survive contact with messy real data are usually built from ranks for this reason.
What are the honest limits?
- The droughts are the strategy's price of admission. Greenblatt warns readers directly in the book: the formula will trail the market for stretches that can run two or three years, and most followers quit during exactly those stretches, which is plausibly why the approach can keep working at all. A backtest you cannot psychologically hold is not an edge you own.
- It is a screen, and screens are blind. The two ratios say nothing about why a company is cheap, whether its earnings are about to fall off a cliff, or whether its industry is in structural decline. Greenblatt's version accepts that blindness deliberately and diversifies across 20 to 30 names. An investor who uses the ranking to concentrate into three favorites has changed the strategy into a different, riskier one while keeping its name.
- The inputs are historical. EBIT is last year's figure. Return on capital describes the business as it was. The formula bets that quality and cheapness persist long enough to be re-rated; for individual companies in fast decline, they do not.
- Published results are backtests and fund history, not a promise. The book presents the formula's historical record, and Greenblatt's firm ran money on related principles. Past mechanics that worked are evidence about the past. Treat anyone's formula, this one included, as a hypothesis your own process still has to grade.
How do investors use it alongside other tools?
Most sensibly as the first pass of a research workflow: the ranking produces the candidate list, and then ordinary due diligence takes over. A stock research checklist run on the top ranks replaces the formula's deliberate blindness where you intend to concentrate. Valuation discipline still applies to whatever survives; a reverse DCF on a top-ranked name tells you what growth its price assumes, which the formula itself never asks.
FullPort's screener filters on the same broad families the formula draws from, earnings strength, returns on equity and capital, and valuation multiples, so a candidate list in the formula's spirit can be built and then extended with your own constraints: start free and build it there. One honesty note: the in-platform magic formula view uses documented proxies for the two ratios (an EV/EBITDA based earnings yield and return on equity), so treat its output as a screening approximation of Greenblatt's ranking, not the published recipe computed to the letter. And a mechanical strategy is the ideal candidate for rehearsal: paper trade it first, through at least one full rotation, and learn whether you can actually hold it before real money grades the lesson.
Educational content, not financial advice. The Magic Formula is a published mechanical strategy described here for understanding. Its historical results do not guarantee or predict future performance, and this explanation is not a recommendation to buy or sell any security.
