This review gives a US investor no evidence that adding Greenblatt's return-on-capital rank pays for itself in a 30-stock portfolio replaced annually. Podgórski gathers plenty of evidence that cheap, profitable stocks outperform local indices. A trader needs an answer to a narrower question. On that question, the review's own sources lean against the formula.
What the formula asks investors to buy
The Magic Formula gives each non-financial, non-utility stock two ranks. Return on capital (ROC) is EBIT divided by net working capital plus net PP&E. Earnings yield (EY) is EBIT divided by enterprise value. Add the ranks, buy the 20 to 30 stocks with the lowest totals at equal weights, and replace the whole portfolio after a year. The proposed source of returns is investors' tendency to overreact to bad news and extrapolate growth in glamour stocks. Podgórski takes this explanation from De Bondt and Thaler and from Lakonishok, Shleifer and Vishny. In factor terms, the combination is value plus profitability. Greenblatt supplies the anchor result: an annual 30.8% from 1988 to 2004, versus 12.4% for the S&P 500.
There is no backtest by Podgórski here. This is a narrative review of post-2006 empirical tests found through Scopus, Web of Science and Google Scholar. Master's and doctoral theses qualify alongside journal articles. The summary table contains nine studies covering the US, the Nordics, Brazil, Hong Kong, Indonesia and a 23-market global panel. Seven rows give a benchmark premium, ranging from 5.63 to 21.80 percentage points. The review concludes that the formula wins in most markets, subject to costs, firm size, market regime and the definitions used for its metrics.
Nine rows, different comparisons
The author is frank about where the evidence comes from: "a significant part of the empirical evidence, especially concerning local markets, originates from unpublished theses or doctoral dissertations". We did not find a count of studies screened or included. The review also leaves unexplained how it selected the nine table entries from the studies discussed.
The comparisons vary considerably. Sistonen reports 29.4% for the Nordics after taxes and costs, against 7.6% for the FTSE Nordic Value Index. Max and Svanberg report 19.2% in Sweden against the broad OMXSPI's 5.7%. In Hong Kong, Fu and Xia's 2.53% a month is for a top-decile portfolio with no index comparison. Most studies use equal weights, which the review acknowledges creates a tilt toward smaller firms. They then measure those portfolios against market indices.
The account of Sasmitapura et al. contradicts itself. In the introduction, their 33.3% CAGR substantially beats the IHSG. In the literature section, the same study fails to beat the IHSG over 2016 to 2020 using LQ45 constituents.
Does ROC improve on earnings yield alone?
The strategy depends on the answer. If EBIT/EV performs as well on its own, ROC amounts to an additional screen.
The review's cleanest comparison finds no benefit from adding it. In Finland from 1991 to 2013, Davydov et al. report annual returns of 19.26% for the complete formula and 13.63% for the market. EBIT/EV alone earns 20.57%. Their assumed transaction cost is 0.5%; the review does not clarify whether the reported returns include it. Podgórski also reports higher risk-adjusted returns for the pure valuation sort.
The US evidence has a similar tilt. Alberg and Seckler examine 1973 to 2012 and find that EY accounts for nearly all the strategy's outperformance during widespread market pessimism. ROC adds relatively little in those periods. In Brazil over 2007 to 2017, Silva et al. find the Piotroski F-score comparable to or better than the formula. The formula returns 11.06%, against 5.04% for the Ibovespa. Blackburn and Cakici then test 23 developed markets over 1991 to 2016. The canonical EBIT version has no statistically significant excess return; it works only when gross profit replaces EBIT.
Podgórski acknowledges the problem. The abstract says strategies based solely on EY "occasionally yield superior results". The conclusion's fourth point also says the original combination of ROC and EY does not appear to hold globally. Still, the author endorses the formula as an imperfect but effective way to exploit inefficiencies. That defence would be persuasive if the winning versions were the formula. One removes its quality leg and keeps its value leg. The other changes the quality metric. Those results support value and gross profitability. They give no such support to Greenblatt's ROC.
Another defence in the review is that ROC distinguishes cheap companies from value traps and accords with Novy-Marx's profitability premium. This is asserted in the review. Davydov is its only cited study comparing the full formula with EY alone, and EY wins there by 20.57% to 19.26%.
Blackburn and Cakici are also enlisted in support of the strategy. The review quotes them saying "large-capitalisation firms contribute significantly to the generated alpha". Their value-weighted alphas exceed equal-weighted alphas everywhere except Europe. We could not establish from the review whether these alphas come from the canonical formula, which the same study finds insignificant, or the gross-profit version. The review never reconciles these uses of the study.
No study in the review compares the full formula with EY alone on matched US portfolios after costs.
The bill for 100% annual turnover
The construction implies roughly 100% turnover a year. Vestre and Wikheim examine Norway from 2003 to 2022, including only firms above NOK 200 million. They report 20.8% against the OSEAX's 12.8%, alongside a statistically significant gross monthly CAPM alpha of 0.50%. The review gives conflicting accounts of what survives costs. One passage says alpha falls "to near zero" and becomes statistically insignificant. Another describes significance "solely at the p < 0.1 level". Reconciling near-zero alpha with significance at 10% is difficult, and the review does not identify the correct account.
Under either description, the table's most careful Norwegian study of returns after costs leaves marginal alpha at best. Sistonen's 29.4% is the table's only return that the review explicitly identifies as net of both costs and taxes. It is measured against a value index. Yet the review treats the formula's persistence as compelling evidence against both the strong and semi-strong forms of the efficient market hypothesis. The evidence in its own cost section cannot sustain that strength of claim.
The formula has been public since 2006. The review's latest sample ends in 2022, in Vestre and Wikheim's Norwegian study, while the author puts post-2020 performance first among topics for future research.
Our US adaptation
We could not trade the international markets covered by the review. Our implementation therefore uses US equities only and says nothing about the Nordic, Brazilian or Asian findings. Each January, we select the 1,000 largest non-ADR US stocks, retain delisted names in the history, and exclude financials and utilities. We sum EBIT over the last four quarterly filings dated strictly before portfolio formation. This substitutes for the three-to-six-month reporting lag recommended in the review.
For ROC, we divide EBIT by current assets minus current liabilities plus net PP&E. For EY, we divide EBIT by market cap plus debt minus cash. We buy the 30 stocks with the lowest combined ranks at equal weights and hold them for twelve months. SPY is the buy-and-hold benchmark. The figures on this page cover the combined formula only; we have no EY-only control here on the same cross-section.
Over the window shown, our run returned 22.85% in total, with a Sharpe of 0.38 and a worst drawdown of -26.19%. Every fill incurs commissions. Two implementation choices warrant doubt before anyone relies on those figures. First, the sector exclusions use a current classification with no date attached. A company reclassified after formation is therefore screened with hindsight. Second, we requested next-day open execution, yet the backtest filled at the close. These returns consequently do not reflect our intended execution schedule.
EY alone wins in the Finnish comparison. Across the 23-market study, the EBIT-based quality rank fails where the gross-profit rank works. Neither result supports retaining Greenblatt's ROC as written.
Our backtest stops at 2024-07-01, and everything after that date is deliberately left untouched so the same strategy can be checked out of sample later.