The Screen

Every month a program sorts about 1,700 US-listed companies from most to least attractive on three measures. That month's money goes to the 15 highest-ranked companies that pass one safety check, skipping any that would take the portfolio past its size limits or that the rules set aside for another reason, such as an agreed takeover, a bankruptcy, or a foreign company the US taxes as a passive investment fund. When a size limit blocks money, it moves on to the next companies down the list. Below is what it measures and why, plus a few things it ends up picking even though nobody asked it to. The exact definitions are in the rules.

Who gets in

Before I score a company, it needs to make it through every one of these gates:

• A US exchange listing. Over-the-counter shares can trade very little, and price, volume, and market value are all read from the US listing.
• A market value above the smallest fifth of NYSE companies (about $1.2 billion in 2026), with $3 million or more of its stock being traded per day. Below that, buying and selling cost more than they should, and the data gets thinner.
• No financial companies or REITs. Their balance sheets work differently, so book value and asset growth mean something else for them.
• No partnerships that send a K-1, and no royalty trusts. A partnership taxed as one sends its owners a K-1 tax form, though one that has chosen to be taxed as a corporation is eligible. A royalty trust collects royalties and pays them out, so it has no business to score.
• Enough history. Three years of annual statements, because the safety check compares this year with last year and last year with the year before.
• Numbers that agree with each other. Some of the screen's numbers come from one data vendor, so I check them. When its market value, share count, currency, or total assets contradict themselves, the company sits out until they don't.

The rules add three smaller gates: a $5 floor for depositary receipts, at least eight quarters of statements, and annual statements no more than 550 days old. When I ran the screen in October, 1,735 companies made it through.

The three scores

Each company is ranked from 0 to 100 on three characteristics, and the three ranks are averaged with equal weight. The idea comes from Eugene Fama and Kenneth French, whose five-factor model¹ accounts for most of the differences in average returns between groups of stocks using five traits. I score three of them, and the other two, the market and company size, come along naturally.

Value: cheap compared with its own industry

Book value divided by market value. A high number means the market prices the company close to what its balance sheet says it owns. Cheap stocks have tended to beat expensive ones over long periods.² Economists haven't settled why that happens. It may be payment for extra risk, or it may be that people avoid companies with dull or troubled stories and overpay for exciting ones.

I compare each company only with its own industry, because a utility and a software company are cheap in different ways. Most of what book-to-market predicts comes from differences between companies in the same industry.³⁴

Market value here is the larger of today's and the one a year earlier (today's share count times the price 12 months back). With today's price alone, a stock that just fell would look cheaper by exactly its fall, and stocks that have been falling have tended to keep falling.⁵ Use only last year's price and a stock that doubled would look just as cheap as it did before the rise. After a broad market fall the month's purchases use today's price instead, because that's when beaten-down stocks have tended to bounce back hardest.⁶

Profitability: what the business earns on what it owns

It's operating profit, before depreciation and after interest, over total assets. Fama and French divide the same profit by book equity,¹ but debt flatters that: a company funded mostly with borrowed money has little equity, so an ordinary profit looks like a large return. Dividing by assets measures what the whole business earns, however it's financed. Two companies can look equally cheap, and the one earning more on what it owns has tended to do better. It pairs well with value:⁷ cheap companies are often cheap because they're struggling, and this score separates the ones that still make money.

Investment: is the company piling on assets?

One-year growth in total assets, where lower is better. Companies that expand quickly, through acquisitions, big building programs, or issuing stock, have tended to deliver worse returns than companies that grow slowly.⁸ The researchers who documented it found risk a poor explanation, and the likelier one is that investors expect too much from fast growers.

Investment and value overlap. Companies that grow their assets slowly also tend to be cheap, so in the Core's factor loadings the value loading can absorb most of the investment effect. Fama and French saw the same overlap in theirs.¹ The holdings still lean hard toward slow growers: in the backtest the companies bought had shrunk their assets by about 6% over the year, when the typical eligible company grew by about 5%.

The trap guard: Piotroski's F score

A cheap stock is sometimes cheap because the business is failing. Before buying, I check whether things are getting better or worse, using a test the accounting professor Joseph Piotroski published in 2000.⁹ Nine yes-or-no questions, a point for every yes. A company scoring 4 or below isn't bought, and neither is one without a score.

Is it making money?

1. Was it profitable? Net income over the assets it started the year with is above zero.
2. Did it bring in cash? Cash flow from operations is above zero. Profit can be an accounting result, and cash is harder to fake.
3. Is profitability improving? Has return on assets improved since last year?
4. Is the profit backed by cash? Operating cash flow is larger than net income. Earnings that run far ahead of cash tend to disappoint later.¹⁰

Is it getting financially stronger?

5. Is debt falling? Debt takes up less of the balance sheet than a year ago.
6. Are its short-term bills easier to cover than they were? Current assets cover current liabilities by more than they did a year ago.
7. Did it avoid issuing shares? The share count didn't rise. A struggling company that sells stock is often raising money it couldn't get elsewhere.

Is it running more efficiently?

8. Are margins improving? Gross profit is a bigger slice of sales than a year ago.
9. Does it squeeze more sales out of what it owns? Sales divided by the assets it started the year with is higher than a year earlier.

The F score is a veto, and it never adds to a company's rank. It measures direction, and the ranking measures level. A company can be cheap, profitable, and slow-growing and still be getting worse fast, and that's the case the guard is for.

Two of the nine are measured a little differently from Piotroski's paper, on purpose. Debt is all debt, where he used long-term debt only: long-term debt drops when a loan comes due within a year and gets relabeled, which looks like repayment when nothing was repaid. And share issuance is read from the share count, because the data provider rarely reports issuance directly.

The line at 4 is one of the three things in these rules I'm least sure about. On five years of backtest, the companies it blocked did slightly better on average than the ones it let through, but the typical company did the same either way. The blocked ones were riskier in both directions: about 5% of them lost more than half their value within a year, against 3% of the ones it let through, and more of them also bounced back hard. Piotroski found his effect mostly among small, thinly followed companies, and every company here is above the NYSE's smallest fifth by market value (about $1.2 billion), so I expect the guard to matter more for avoiding disasters than for raising returns. The record will show how that works out.

What it picks up without trying

Choosing on three characteristics pulls in others. On the October list, compared with every eligible company:

• Smaller companies. The market value floor and buying toward equal position sizes tilt the Core toward small and mid-size companies. The backtest's size loading was about 0.7, the strongest of any factor besides the market, with value close behind.
• Buybacks. A company that buys back stock with its cash shrinks its assets, which both the profitability and investment scores reward.
• Dividends. Profitable, slow-growing companies have cash to hand back.
• Companies based abroad. Sanofi, Shell, and others listed in the US through depositary receipts. They bring currency risk and foreign withholding tax on dividends. The US is capped at its share of world stock markets (62% for now), so in the backtest from 2016 a median of 38% of the portfolio was based outside the US, more than the list's 27% would suggest.
• Momentum, a slight lean against. With value on the larger market value, the backtest's top 15 had typically risen about 3% over the previous year, while the pool rose 9%. A rise makes a stock look more expensive at once, and a fall makes it look cheaper only a year later, so the ranking leans a little away from recent winners. On today's price alone they'd typically have fallen about 5% while the pool rose, a stronger lean toward stocks that keep falling. Because of the rule that holds winners, strong risers stay in a little longer. No other rule here treats a rising price as a point in a stock's favor. A stock other investors have bid up is hardly one they avoid, so this rule rests on the evidence alone: past-year winners have tended to keep rising. All it does is delay a sale, and it never picks what I buy. In the backtest from 2016, winners below their sell line were about 3% of the portfolio on average and 9% at the most.

One tilt I expected is slight. Buyable companies carry a little less debt than the rest, relative to their assets (liabilities of about half their assets, against 55%), so "cheap" here still doesn't mean a fortress balance sheet. Taken one at a time, they're about as volatile as the market. Together they swing more than it: in the backtest from 2016 the portfolio moved about 1.27 times as much as VT and fell further in sell-offs, -51% at its worst against VT's -34%. Funds built on the same tilts show the same thing.

What it never looks at

The screen never uses forecasts, analyst ratings, or estimates of fair value, and it never reads the news. It uses only what companies have already reported and the prices their stocks have closed at: today's, and a year ago's for value and momentum. Every input is a number anyone can check.

The honest limits

Averaged over decades, these characteristics have paid off. They've also gone missing for a decade at a time, and value's dry spell from about 2007 to 2020 is the best-known example. Nothing here promises they'll pay during my record. The backtest runs ten years, still a single path, and I tuned the rules on it. Funds that run the same tilts with real money, such as Dimensional's small-cap value funds, beat the total US market by a point or two a year over their 25 to 33 years, and trailed it by about two and a half points a year from 2016. The part the record can show is whether I keep following the rules while those traits are out of favor.

Sources

1. Eugene F. Fama and Kenneth R. French, "A Five-Factor Asset Pricing Model," Journal of Financial Economics 116, no. 1 (2015): 1-22. ↑
2. Eugene F. Fama and Kenneth R. French, "The Cross-Section of Expected Stock Returns," Journal of Finance 47, no. 2 (1992): 427-465. ↑
3. Clifford S. Asness, R. Burt Porter, and Ross L. Stevens, "Predicting Stock Returns Using Industry-Relative Firm Characteristics," AQR working paper (February 2000). ↑
4. Randolph B. Cohen, Christopher Polk, and Tuomo Vuolteenaho, "The Value Spread," Journal of Finance 58, no. 2 (2003): 609-641. ↑
5. Narasimhan Jegadeesh and Sheridan Titman, "Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency," Journal of Finance 48, no. 1 (1993): 65-91. ↑
6. Kent Daniel and Tobias J. Moskowitz, "Momentum Crashes," Journal of Financial Economics 122, no. 2 (2016): 221-247. ↑
7. Robert Novy-Marx, "The Other Side of Value: The Gross Profitability Premium," Journal of Financial Economics 108, no. 1 (2013): 1-28. ↑
8. Michael J. Cooper, Huseyin Gulen, and Michael J. Schill, "Asset Growth and the Cross-Section of Stock Returns," Journal of Finance 63, no. 4 (2008): 1609-1651. ↑
9. Joseph D. Piotroski, "Value Investing: The Use of Historical Financial Statement Information to Separate Winners from Losers," Journal of Accounting Research 38, supplement (2000): 1-41. ↑
10. Richard G. Sloan, "Do Stock Prices Fully Reflect Information in Accruals and Cash Flows about Future Earnings?" The Accounting Review 71, no. 3 (1996): 289-315. ↑