28 years · 25,050 point-in-time valuations · 1998–2025
Over 28 years, the stocks our engine called cheap turned $10,000 into about $14,470 in a typical five-year hold.
The ones it called expensive turned the same $10,000 into about $13,140.
Called cheap
$14,470
+44.7% typical five-year total return
Called expensive
$13,140
+31.4% typical five-year total return
“Typical” means the median — half the picks did better, half worse. On averages the same comparison is +66.0% against +48.3%. The dollar figures restate those five-year total returns on a $10,000 stake: they describe one five-year hold, not 28 years of compounding, and they are hypothetical study results rather than a track record.
28
years replayed
1998–2025, every June
25,050
point-in-time valuations
the 1,500 biggest US companies — including the ones that later failed
+13.3pp
typical five-year gap
between what it called cheap and what it called expensive
The cohorts
Cheap beat expensive at every holding period
Typical (median) total return, dividends included, for each of the three buckets the engine sorts stocks into. The order never flips — and the gap widens the longer you hold.
- Cheap
- About right
- Expensive
After 1 year
After 3 years
After 5 years
Against the index
Versus the S&P 500
The cheap bucket's five-year return minus the official S&P 500 Total Return index, by era — the part of our own results we like least.
the centre line is the S&P 500 Total Return
It works when the market is rational about price — the dot-com era was its finest hour, and it roughly kept pace through the financial-crisis years. It lags when a handful of giants carry the index, as they did right through the growth decade.
The honest headline
Cheap picks beat expensive picks against the index on every single measure — but over five years the typical cheap pick still finished −13.1pp behind the S&P 500, and only 44% of them beat it. What this study proves is that the engine ranks stocks against each other. It does not prove — and we do not claim — that any individual pick beats the index.
| Group | 3 years | Beat it | 5 years | Beat it |
|---|---|---|---|---|
| Cheap | +3.2pp / −4.8pp | 46% | +1.3pp / −13.1pp | 44% |
| Expensive | −3.4pp / −9.9pp | 42% | −9.2pp / −21.3pp | 39% |
Everything
All of the numbers
Total return after 1, 3 and 5 years, dividends included. Each cell is the average first, then the typical (median) result.
| What the engine said | After 1 year | After 3 years | After 5 years |
|---|---|---|---|
| Cheap (15%+ below our fair value) | +13.0% / +11.3% | +38.4% / +29.4% | +66.0% / +44.7% |
| About right (0–15% below) | +10.8% / +10.5% | +34.2% / +28.5% | +58.9% / +42.7% |
| Expensive (above our fair value) | +9.4% / +6.7% | +27.5% / +20.1% | +48.3% / +31.4% |
| Signal strength | Real at 1, 3 and 5 years (95% confidence) | ||
Signal strength is the statistical test of whether the engine's ranking lines up with what actually happened next. It clears the 95% confidence bar at every holding period — and it survived the data correction described in the methodology.
The verdict
Is the result good?
Yes — and here is exactly how good. The engine's ranking held up across 28 years and 25,050 valuations of the 1,500 biggest US companies, counting every one that later went bankrupt, was bought, or quietly stopped trading. Cheap beat about-right, and about-right beat expensive, at every holding period, on both the average and the typical stock. That is not a lucky decade; it is nearly three of them.
And here is what it is not. It is not a promise to beat the index, and it is not steady. Across the 2014–2020 starting dates cheap names trailed the S&P 500 by −24.4pp on average and only 29% of them beat it — the better part of a decade in which this discipline simply did not pay. Anyone using it should plan on sitting through a stretch like that, because there is no version of value investing that skips them.
The long version
How we ran it — methodology, corrections, and what we DIDN'T claim →
How the test was built, what we corrected along the way, the limits of what it proves, and every caveat in long form.
Hypothetical backtested results, shown for research and education — not investment advice and not a promise of future performance. Past results, real or simulated, do not predict future results. Figures are gross of fees, taxes and trading costs; groups are equal-weighted; delisted names are carried at their final adjusted close. The benchmark is the official S&P 500 Total Return index (^SP500TR). Full methodology, corrections and caveats: read the long version.