The classic value screen returns 49 names. 21 are financials

Almost every free stock screener ships with the same starter filters. Price to earnings under 15. Return on equity over 15%. Debt to equity under one. It is the textbook value screen, it is genuinely sensible, and it has been in print since Benjamin Graham.

I ran it across our universe of 975 companies, using each company's most recently filed annual report. Forty-nine survive.

That is the first problem, and it is not the one people expect.

The list is not bad. That is what makes it dangerous

The obvious criticism of a simple screen is that it finds rubbish. It does not.

Those 49 companies have a median return on invested capital of 17.5%, against 9.8% for the universe as a whole. Only four of them sit below the universe median. By the measure that actually tracks business quality, the screen is picking well above average.

So this is not a story about a broken tool. It is a story about a tool that answers a slightly different question than the one you thought you asked, and returns a plausible enough list that you never notice.

Forty-nine names, and nearly half are one sector

Sort the survivors by sector and the shape appears immediately.

Twenty-one of forty-nine. Banks and insurers are 43% of the output.

This is not because banks and insurers were unusually cheap that year. It is structural. Lending and underwriting are businesses that run on other people's money, so a healthy insurer naturally carries a high return on equity, and the sector has traded on low earnings multiples for most of the last two decades for reasons that have nothing to do with any individual company being overlooked.

Put a low-multiple filter and a high-ROE filter in the same screen and you have not built a quality filter. You have built a financials filter, and it does not say so anywhere on the label. If you run that screen every month and buy the top names, you are making a large, undiversified sector bet you never decided to make.

The ROE half of that is the same mechanism I wrote about in return on equity being moveable by the balance sheet: the ratio rewards borrowed money, and financials run on more of it than anyone.

What it throws away is worse than what it keeps

Ninety companies in the universe earn a top-decile return on invested capital, which here means 28.9% or better. These are, by that measure, the best businesses available.

The screen admits nine of them.

Thirty-six are knocked out by the price-to-earnings filter alone. They pass the profitability test and they pass the debt test. They fail on the multiple and nothing else.

Another forty fail the debt test, and that is its own quiet story. A company that has spent years buying back its own stock has less equity left to show against its borrowings, so the debt-to-equity test reads it as risky. That is the same mechanism that gave Home Depot a 110% return on equity: the buybacks that flatter one ratio trip another.

That is the whole trap in one number. A business earning 30% or 40% on the capital inside it is not usually available at twelve times earnings, because other people can also see the return it makes. Filtering those out is not prudence. It is a rule that removes the best companies from consideration on every single run, structurally, for the entire period during which owning them would have mattered.

And the exclusion is permanent in a way that is easy to miss. A company you never see in a list is not a decision you made and got wrong. It is a decision you never made at all.

Why everyone's list looks the same

The last problem is the quiet one. These are default filters. They are the defaults on most free tools, they are the numbers in the textbooks, and the universe of large listed companies is the same for everybody.

So the output is the same for everybody too.

Whatever edge existed in a list of statistically cheap companies with decent returns was competed away a long time ago, because the list is not scarce. It costs nothing and takes four seconds. Anything that easy to compute is already in the price.

That is not an argument against screening. Screening is how you get from thousands of companies to a number you can actually read. It is an argument against thinking the screen is the work.

A better way to use the same tool

None of this requires abandoning the screener. It requires demoting it. Three changes do most of the work.

Invert it once. Run the same filters backwards: price to earnings above 25, return on invested capital in the top decile. That list is everything the classic screen structurally refuses to show you. You do not have to buy any of it. You just have to know what you are turning down, because right now that refusal happens silently, on every run.

Loosen one filter at a time. Drop the debt rule and watch who arrives. If the newcomers are companies that spent a decade buying back stock while earning high returns on capital, the rule was screening out balance sheet policy, not risk. If they are fragile businesses carrying debt they cannot service, the rule was doing its job. Either way you learn what the filter was actually doing, which the filter itself never tells you.

Treat the output as a reading list, not a shopping list. Forty-nine names is not a portfolio. It is a queue. A screen's one legitimate job is to decide what you read next, and the moment its output skips the reading and goes straight into an order form, the decision has been handed to the arithmetic of whatever defaults the tool shipped with. The people who chose those defaults never met your portfolio. That is not a criticism of them. It is a reason not to let their settings decide what you own.

What we do differently, and what we do not

We screen too. Everyone does. The differences that matter are these.

We start from return on invested capital rather than return on equity, so the ranking does not quietly reward balance sheet decisions.

We cap sector concentration rather than letting the arithmetic pick a sector for us, which is what the 21 financials above are.

We treat a low multiple as a claim that needs checking, not as a discount, for the reasons in why cheap is a forecast.

And we publish which pillars go into the score without publishing how they are weighed. The methodology has the shape of it. The weights are the part we keep, and I would rather say that plainly than pretend there is nothing behind the curtain.

---

Screen run on 9 September 2026 against our own annual financials table, using each company's most recently filed annual report rather than a fixed calendar year. Universe: 975 companies. Sector labels from our own securities table.