The short answer
- Five ways of picking were compared across four tests (two periods each in Japan and the US). Only 'EV/EBIT + F-score 7+' beat the market (the median of all companies) in all four, by +3% to +26%.
- Greenblatt's magic formula (cheapness + return on capital) beat the market in 3 of four and lagged in 1. Cheapness alone beat it in 3 and lagged in 1 (US first period: -2%).
- We also checked whether cheapness simply picks poor businesses. In Japan, the companies picked on cheapness alone had a return on capital of 22%, above the 16% of all companies, and 1% of them fell 30% or more (all: 9%).
- But both Japanese periods fall within the years when the stock exchange's push on companies trading below book value lifted cheap stocks. The US top 30 alone were uneven (they lagged the market in the first period).
Five methods, four tests
Each method ranks the companies on figures known at the time; we then look at the top fifth.
| Method | Japan period 1 | Japan period 2 | US period 1 | US period 2 |
|---|---|---|---|---|
| EV/EBIT + F-score 7+ (this site's method) | +26%+64% | +3%+25% | +10%+9% | +6%+14% |
| Cheapness only (EV/EBIT) | +30%+68% | +6%+28% | -2%-3% | +6%+15% |
| Magic formula (cheapness + return on capital) | +22%+60% | -2%+21% | +16%+16% | +3%+11% |
| Magic formula + F-score 6+ | +16%+54% | -1%+22% | +16%+16% | -0%+9% |
| Cheapness + return on capital above median | +16%+54% | -1%+22% | +15%+14% | +5%+13% |
Large figure: the top fifth’s median return minus the median of all companies in the same test. Small figure: the top fifth’s median return. Japan: companies worth ¥30bn or more; US: $2bn or more.
- Japan: 2021–22 annual reports → 3 years
- Japan: 2023–24 annual reports → 23 months
- US: FY2021 results → 3 years from Apr 2022
- US: FY2023 results → 29 months from Apr 2024
Does cheapness pick poor businesses?
The median return on capital and F-score of the companies each method picked (top fifth), and how many fell 30% or more.
Japan: 2021–22 annual reports → 3 years
| Return on capital | F-score | Fell 30%+ | Median return | |
|---|---|---|---|---|
| EV/EBIT + F-score 7+ (this site's method) | 20% | 8.0 | 1% | +64% |
| Cheapness only (EV/EBIT) | 22% | 7.0 | 1% | +68% |
| Magic formula (cheapness + return on capital) | 38% | 7.0 | 2% | +60% |
| All companies | 16% | 7.0 | 9% | +37% |
Japan: 2023–24 annual reports → 23 months
| Return on capital | F-score | Fell 30%+ | Median return | |
|---|---|---|---|---|
| EV/EBIT + F-score 7+ (this site's method) | 19% | 8.0 | 3% | +25% |
| Cheapness only (EV/EBIT) | 24% | 7.0 | 2% | +28% |
| Magic formula (cheapness + return on capital) | 38% | 7.0 | 4% | +21% |
| All companies | 15% | 7.0 | 7% | +23% |
US: FY2021 results → 3 years from Apr 2022
| Return on capital | F-score | Fell 30%+ | Median return | |
|---|---|---|---|---|
| EV/EBIT + F-score 7+ (this site's method) | 56% | 7.9 | 22% | +9% |
| Cheapness only (EV/EBIT) | 52% | 7.0 | 28% | -3% |
| Magic formula (cheapness + return on capital) | 118% | 6.8 | 24% | +16% |
| All companies | 38% | 6.0 | 29% | -0% |
US: FY2023 results → 29 months from Apr 2024
| Return on capital | F-score | Fell 30%+ | Median return | |
|---|---|---|---|---|
| EV/EBIT + F-score 7+ (this site's method) | 51% | 7.9 | 14% | +14% |
| Cheapness only (EV/EBIT) | 39% | 6.8 | 20% | +15% |
| Magic formula (cheapness + return on capital) | 90% | 6.8 | 20% | +11% |
| All companies | 34% | 6.0 | 22% | +9% |
Limits
Japan: companies delisted other than by tender offer could not be followed (those bought by tender offer count at the offer price). US: companies no longer in the SEC ticker list (acquired or bankrupt) are left out. Only two periods each, and we adopted the best of the methods compared, so results may look better than they will be.
This site ranks companies whose share price is low relative to their profits, using public filings and market prices, and has AI draft hypotheses on why they are cheap from those filings. The hypotheses are unverified. Nothing here is a price forecast or a recommendation to buy or sell. Not investment advice.
Updated 29 Sep 2026 12:31 JST