The same idea, read one way, was our most reliable winner. Read the other way, it was this. The teardown sits them side by side.

The overview showed the ranking working perfectly and backwards. The teardown answers the obvious next question: if the clean names lose, does buying fewer of them, the very cleanest, make it worse? It does, and that tells you what this screen really is.

The purer you made it, the more it lost

Hold the broad basket and you got about 8.7% a year. Concentrate into the purest ten, the companies with the most immaculate cash conversion, and the return fell further, toward 8%, through an even deeper drawdown near 59%. A true ranking rewards concentration at its best end. This one punished it, because its best end is its worst stocks. Every step toward "higher quality" was a step down in return.

The same measure, the opposite trade

Here is the resolution. The accruals screen uses this exact profit-versus-cash gap, but as a way to avoid the companies whose profits are least backed by cash, the likely manipulators and stumblers. Used as an exclusion, it worked. This screen takes the same measure and uses it to select, buying the most cash-backed names outright, and in doing so it loaded up on finished, slow-growing cash cows. The signal was never wrong. The instruction attached to it was. Avoid the worst, and you sidestep ruin. Buy the "best", and you inherit a decade of low-growth compounding.

No regime rescued it

You might hope it at least shone in some particular weather. It did not. Split the record by growth, inflation and rates and it stayed a step behind almost everywhere, positive in level but rarely ahead of a plain index fund. There is no macro backdrop this version was built for, because the problem is structural, not cyclical.

How often did buying clean actually beat the market?

Less than half the time over one and three years, and only late in the sample did the five year runs pull convincingly ahead, on the back of small, cheap names doing the work rather than the clean earnings themselves.

What it held, and the worst it would have handed you

Two last checks make the risk concrete: what the basket was actually made of, by company size, and, instead of the single drawdown path, the worst fall you would have suffered no matter which month you started and held for five years.

Keep it as a filter, not a pick

The honest use of cash-backed earnings is the one the accruals screen pointed to: as a red flag, a reason to drop a company whose profits refuse to become cash, not a reason to buy the ones that already have. Read that way, the same numbers on this page become useful again. That is the whole argument for checking rather than trusting. Nobody would guess, from the textbook, that "buy the highest quality earnings" was a decade-long mistake in India, and nobody would keep believing it after watching this. The measure is a keeper. The direction was the error.