Value investing: what it is and its S&P 500 record

Benjamin Graham and David Dodd set out the idea in Security Analysis (1934), and Graham popularized it in The Intelligent Investor (1949): a stock's price and the worth of the business behind it do not always match, and buying when the price is well below leaves a margin of safety. Decades later, academic research — the most cited paper is Fama and French's, from 1992 — documented that stocks cheap relative to their book value had outperformed expensive ones.

There are two ways to do it. The analyst's, company by company, estimating what each business is worth. And the quantitative one, this site's: a rule that measures how cheap each of the ~500 companies is with a few ratios, the same for all of them every month. The first can tell a struggling business from an undervalued one; the second cannot, and in exchange it never falls in love with a company.

Value investing is not buying what has fallen: a stock down 40% can still be expensive if its earnings fell further. And a ratio saying “cheap” does not mean the market is wrong.

They are used upside down compared with the familiar ratios for a practical reason: a loss-making company has a negative P/E that means nothing, while its negative earnings yield can be ranked with the others. The higher the yield, the cheaper the company.

Each component is compared with the rest of the index (a Z-score, clipped at ±3 so that one extreme case does not dominate), the components are averaged and the companies ranked. The portfolio keeps the top ones, in equal parts, and is rebuilt every month. The accounts come from the SEC and are used only from the day they were filed: in 2005, the engine does not know what was filed in 2006.

Broken data are dropped, not clipped: applying Berkshire Hathaway's class A EPS to its class B gave an earnings yield of 4,253% and put it first in the index on value.

The full recipe, with each component's limits, is in the methodology. Today's order is in the rankings, and every company page — for example, Apple's — shows the company's value score next to the other three factors. On September 18, 2026, 494 of the 503 companies on the list had a value score.

The site publishes the rule in four sizes. Split in equal parts, the fair comparison is the S&P 500 also in equal parts:

All four end above the index, and all four fell far more: between 75% and 82%, against the index's 60%. That is not a detail. An 80% drawdown needs what is left to grow fivefold just to get back to the start.

Weighted by company size, like the classic S&P 500, the yardstick is SPY, the fund that tracks that index, which returned 8.68% a year with dividends:

It wins in five periods out of eight, but loses by a wide margin in the two that hurt most. In the dot-com bubble, when expensive stocks were the ones rising, the portfolio lost money while the index stayed flat: on March 2, 2000, every 100 dollars from the start was worth 79. And in the financial crisis it fell much more than the index. That is the risk known as a value trap (see the frequently asked questions): what looks cheap can get much cheaper.

The worst run, measured with the curve's monthly points (which always fall somewhat short of the daily ones): from June 11, 2007 to March 11, 2009, the portfolio lost 76.95%. That day it was worth 39% less than ten years earlier, when it started, and it did not regain its peak until February 11, 2013. The month after the bottom it rose 70.9%, and that rebound falls inside the “growth decade” period, which starts in February 2009: counted from April 9, 2009, after the rebound, its lead in that period drops from 6.2 to 4.4 points a year. Its worst month was a different one: from February 13 to March 16, 2020, −48.34%.

The standard test is the information coefficient (IC): each month, the rank correlation between the value score and what each stock did the following month. Over 331 months, with every company that passed through the index:

The interesting part is the ladder. Split the index into ten groups by value score, and the cheapest do better than the next, which do better than the next, in almost perfect order (+0.90 out of a maximum of 1). It is the only one of the site's four factors with that ladder; over twelve months, the cheapest decile beats the most expensive by 3.4 points. But the IC is one hundredth and the IC-IR 0.06: the signal points the right way and is very small next to the noise of any given month.

In exchange, it trades little: the 20-stock value portfolio turns over 177% of its positions a year, against 431% for momentum (a 2021-2026 measurement, good for the order of magnitude).

What is the difference between value and growth investing?

Value means holding the companies that are cheap relative to what they earn, own or sell; growth, the ones whose sales and earnings grow fastest, even if they trade expensively. In the 1998-2026 simulation, the 20-stock growth portfolio would have returned 7.93% a year and the value one 11.36%, against 10.37% for the equal-weight S&P 500.

Does a low P/E mean a stock is cheap?

Not necessarily: last year's earnings may be about to fall. That is why this site does not look at the P/E alone but at three measures (earnings, book value and sales), and requires at least two.

What is a value trap?

A company that looks cheap because its business is deteriorating, and keeps getting cheaper. A quantitative rule cannot tell them apart. The three deepest drawdowns among the site's twenty equal-weight portfolios all belong to value: the 10, the 20 and the 50.

Has value investing stopped working?

These data can neither confirm nor deny it. In the simulation, the 20-stock value portfolio beat the equal-weight index by one point a year and won in five of eight periods, but with more risk, and its alpha is negative. The signal points the right way, but it is small and not proven.

Where can I see an S&P 500 company's value score?

In the day's rankings and on each company page (for example, Apple's). The score describes; it is not a buy recommendation.

• They are simulated. Nobody traded this portfolio in 1998. It is the rule applied forward, month by month, with the data that existed at each date.
• They carry no costs. No commissions, no spread, no taxes, no currency conversion. All figures are in dollars.
• Survivorship bias: nearly closed, not fully. The history includes the prices of companies that later left the index; around 0.1% of membership slots are missing.
• It is one definition of value among many. Other ratios or another portfolio size would give another number, and a manager picking company by company does not do what this rule does.
• It is a single market. The 500 largest US companies over 27.6 years. What happened here need not happen elsewhere or in another era.

The portfolio, its curve, its holdings and its monthly ins and outs are on the value 20 page; the curve downloads as CSV, no account needed. To try another size or weighting, use the simulator. All twenty portfolios side by side are in Model Portfolios.

More guides: what momentum is, three ways to invest in the S&P 500, how to read a portfolio's Sharpe ratio and drawdown and what insider buying is.