While the Magic Formula appears to work over the long-term, Greenblatt discusses reasons why it is not used by everybody. He considers it good news that the formula doesn't work in many periods, as this prevents people from exploiting it and thus reducing its returns.
For example, the formula is outperformed by the market in 5 months out of every 12 (for the 17 year period Greenblatt tested). It can even underperform for years in a row, as at one point in the simulation it underperformed for three years in a row. This makes it very difficult for fund managers to employ it consistently. Greenblatt cites examples of various value fund managers he knows who lost large chunks of clients or who even closed down shop after underperforming for years, only to eventually emerge with far superior returns in the long run. The pressure on fund managers not to underperform their peers forces them to make decisions that are not in the best interests of their funds in the long term.
But to be able to stick to the formula, Greenblatt argues that investors must understand why it works. The bottom line is that by employing the formula, investors are buying stocks with the best combination of returns on capital and earnings yield. High returns on capital suggest a business has an advantage or position which allows it to make abnormally high profits, and a high earnings yield means investors are paying relatively less to buy each dollar of earnings. This combination is logical, and so it makes sense that the formula would work in the long-term.