Does Greenblatt's Magic Formula still work?
Updated October 28, 2026
On QuantGPT's test from 1998 to September 2026, the library's Greenblatt-inspired Magic Formula run beat the S&P 500 with dividends on both growth and Sharpe, but fell behind it on both after the walk-forward split, which QuantGPT labels TRADE-OFF. It grew 15.4% a year against 9.4%, gross of trading costs, with a Sharpe ratio of 0.72 against 0.66.
Greenblatt ranks on operating earnings over enterprise value plus return on capital and holds each stock about a year. This run ranks on net income over market value among stocks with a return on equity of 15% or more, re-formed monthly, so it tests a cousin of the book's screen, not the screen itself.
The original claim
Joel Greenblatt, 'The Little Book That Beats the Market' (2005). Rank the market on cheapness (earnings yield) plus quality (return on capital); buy the best combined ranks mechanically.
In plain words: leave out utilities, banks and other financial firms, foreign stocks and the smallest companies. Rank the rest on how cheap each is (operating earnings over enterprise value) and on how good a business it is (operating earnings over the capital it uses). Add the two ranks, buy the 20 to 30 best a few at a time over the year, and hold each for about a year.
- Earnings yield = EBIT / EV; return on capital = EBIT / (net working capital + net fixed assets)
- Exclude utilities, financials and foreign stocks; apply a market-cap floor
- Sum the two ranks; buy the top 20-30 in batches; hold each about one year
Book backtest: 30.8%/yr vs 12.4% for the S&P 500, 1988-2004, on the largest 3,500 US stocks. Independent tests (e.g. Gray and Carlisle, 2012) find it beat the market by less, with EBIT/EV alone doing better and multi-year droughts.
How QuantGPT's run differs
Holds the 30 liquid US stocks with the highest earnings yield (net income over market value) among those with a trailing return on equity of at least 15%, equal weight, long only, re-formed monthly. Greenblatt summed the ranks of EBIT over enterprise value and EBIT over net working capital plus net fixed assets, left out financials and utilities, used the largest 3,500 US stocks and held each about a year, 1988 to 2004. Return on equity is levered and net income is after interest and tax, so both legs differ.
- Universe: liquid US stocks: raw close $5 or more, the top 3,000 by 60-day median dollar volume, 5M shares where the count is known.
- Benchmark: the S&P 500 with dividends reinvested ($SPXTR), bought and held.
- Cash: a month with fewer than 15 qualifying stocks is held in cash. No month was: the first book formed at the end of January 1998, and every one of the 344 months held stocks.
- The engine's own notes: 1 held position had a period return outside -95% to +200% and was cut to the bound (a guard against bad prints). If a cut return is real, the book's return, volatility and Sharpe differ from what the move would give. Delisted or halted holdings stay in the book: they exit at the last trade inside the holding period, and a holding with no trade at all after formation books -30% (Shumway 1997).
- In short: Ranks on earnings yield (net income over market cap) with an ROE floor; Greenblatt sums ranks of EBIT/EV and return on capital.
The result
Over this window the criteria beat the benchmark on both growth and Sharpe, but fell behind it on both after the walk-forward split, which QuantGPT labels TRADE-OFF: a description of what the data did, not a prediction and not a recommendation.
| Strategy | S&P 500 with dividends | |
|---|---|---|
| Growth a year, 1998 to September 2026 | 15.4% | 9.4% |
| Worst fall from a peak | -60.9% | -50.9% |
| Sharpe ratio | 0.72 | 0.66 |
| Volatility a year | 24.1% | 15.4% |
| Share of winning months | 61.9% | 64.5% |
| Growth a year, before 2015 | 18.1% | 6.2% |
| Sharpe ratio, before 2015 | 0.81 | 0.47 |
| Growth a year, from 2015 | 11.6% | 14.2% |
| Sharpe ratio, from 2015 | 0.59 | 0.97 |
| Net of 25 bps a side: growth a year | 14.1% | not modelled (bought and held) |
| Net of 25 bps a side: Sharpe ratio | 0.67 | |
| Net of 25 bps a side: worst fall | -61.5% |
| Cost a side | Growth a year | Sharpe ratio | Worst fall | Cost a year |
|---|---|---|---|---|
| 5 bps | 15.2% | 0.71 | -61.0% | 26 bps |
| 10 bps | 14.9% | 0.70 | -61.2% | 51 bps |
| 25 bps | 14.1% | 0.67 | -61.5% | 128 bps |
| 50 bps | 12.9% | 0.63 | -62.3% | 255 bps |
Every year
Losing years for the strategy: 2008 (-42.2%), 2011 (-17.5%), 2015 (-11.0%), 2018 (-14.3%), 2020 (-5.2%), 2022 (-15.5%).
| Year | Strategy | S&P 500 with dividends |
|---|---|---|
| 1998 February to December | 14.2% | 27.2% |
| 1999 | 8.6% | 21.0% |
| 2000 | 31.2% | -9.1% |
| 2001 | 48.5% | -11.9% |
| 2002 | 9.4% | -22.1% |
| 2003 | 72.2% | 28.7% |
| 2004 | 46.0% | 10.9% |
| 2005 | 11.9% | 4.9% |
| 2006 | 25.7% | 15.8% |
| 2007 | 5.8% | 5.5% |
| 2008 | -42.2% | -37.0% |
| 2009 | 48.9% | 26.5% |
| 2010 | 37.4% | 15.1% |
| 2011 | -17.5% | 2.1% |
| 2012 | 15.3% | 16.0% |
| 2013 | 38.8% | 32.4% |
| 2014 | 15.0% | 13.7% |
| 2015 | -11.0% | 1.4% |
| 2016 | 31.8% | 12.0% |
| 2017 | 29.1% | 21.8% |
| 2018 | -14.3% | -4.4% |
| 2019 | 24.9% | 31.5% |
| 2020 | -5.2% | 18.4% |
| 2021 | 33.6% | 28.7% |
| 2022 | -15.5% | -18.1% |
| 2023 | 30.2% | 26.3% |
| 2024 | 20.6% | 25.0% |
| 2025 | 23.0% | 17.9% |
| 2026 January to September | 2.5% | 12.7% |
Run it yourself
Open it in the terminal from the strategy library (RUN IT on its entry), or read the public result page at terminal.quantgpt.co/study/greenblatt-magic-2005. Change the start year, the cutoffs, the number of stocks or the cadence and run it again. Or ask your AI, connected over MCP:
Use QuantGPT to backtest the Magic Formula from 1998, with costs at 25 bps a side, and show the losing years.
Questions
Does the Magic Formula still work?
On QuantGPT's test it beat the market over the whole window and trailed it after 2015. The library's Greenblatt-inspired run, the 30 liquid US stocks with the highest earnings yield among those with a return on equity of 15% or more, re-formed monthly, grew 15.4% a year from 1998 to September 2026, gross of costs, against 9.4% for the S&P 500 with dividends, with a Sharpe of 0.72 against 0.66. Before 2015 it grew 18.1% a year against 6.2%; from 2015, 11.6% against 14.2%. Over this window the criteria beat the benchmark on both growth and Sharpe, but fell behind it on both after the walk-forward split, which QuantGPT labels TRADE-OFF. The run ranks on net income over market value, a cousin of Greenblatt's EBIT measures, so it is not the book's own screen.
What is Greenblatt's Magic Formula?
Joel Greenblatt's The Little Book That Beats the Market (2005) ranks companies on two measures: earnings yield, operating earnings (EBIT) over enterprise value, and return on capital, EBIT over net working capital plus net fixed assets. Add the two ranks, buy the 20 to 30 best a few at a time, hold each about a year, and leave out utilities, financial firms and foreign stocks. The book reports 30.8% a year against 12.4% for the S&P 500 from 1988 to 2004, on the largest 3,500 US stocks.
What were the Magic Formula's worst years?
In QuantGPT's run the losing calendar years were 2008 (-42.2%), 2011 (-17.5%), 2015 (-11.0%), 2018 (-14.3%), 2020 (-5.2%), 2022 (-15.5%). Its worst fall from a peak, at month-end closes, was -60.9% against -50.9% for the S&P 500 with dividends. All gross of trading costs; net of 25 bps a side it grew 14.1% a year.
Which value investing strategies still beat the market?
QuantGPT keeps no ranking of them. Each of the 55 runnable entries in its strategy library can be run on the same data, and every run splits its statistics at a cut year (2015 unless you pick another), so the years after the cut read on their own. On that test the Magic Formula run trailed the S&P 500 with dividends after 2015: 11.6% a year against 14.2%, gross of costs.
Where can I backtest the Magic Formula?
On QuantGPT: from the strategy library with one click (RUN IT on the Magic Formula entry opens it in the terminal), or from Claude or ChatGPT connected over MCP, with a question such as: Use QuantGPT to backtest the Magic Formula from 1998, with costs at 25 bps a side, and show the losing years. You can change the start year, the return on equity floor, the ranking, the number of stocks and the cadence. QuantGPT's library documents 175 strategies, 55 of them runnable in one click on decades of US market history. QuantGPT is $49 a month and QuantGPT Pro $149, each with a 3-day free trial.
Related
- Strategy libraryDoes 12-1 momentum still work?
- Strategy libraryDoes post-earnings announcement drift still work?
- Strategy libraryDoes the Piotroski F-Score still work?
- ComputeMCP server for backtesting
- DataThe research warehouse
Backtests show past results, gross of trading costs unless stated. A backtest is not a promise. QuantGPT is research software and nothing here is investment advice. The disclosures and terms govern.