Magic Formula

The Magic Formula ranks every stock twice — once on earnings yield (EBIT / enterprise value) and once on return on capital (EBIT / (net working capital + net fixed assets)) — then adds the two ranks and buys the best combined scores. Joel Greenblatt published it in The Little Book That Beats the Market (2005) as a mechanical way to buy good businesses (high return on capital) at cheap prices (high earnings yield).

The mechanics are strict: hold 20-30 qualifying stocks, replace them after a year, exclude financials and utilities, and repeat. Greenblatt's backtest claimed roughly 30% annualized from 1988 to 2004, though independent replications since publication have found smaller edges, and the strategy has had multi-year stretches of trailing the index.

Greenblatt's own explanation for why it keeps working is that it periodically does not — the formula underperforms often enough, sometimes for two or three years straight, that most people abandon it before the averages assert themselves. A backtest is not a guarantee; the behavioral cost of holding through the bad stretch is the real price of admission.