It bought the companies with the cleanest, most honest profits. The ranking sorted almost perfectly, and put them at the bottom.

Each year you bought the thirty companies whose profits were most fully backed by cash, exactly what a careful analyst would call high quality. Held straight through they lagged, and through the deepest drawdown on this page. Across every start month, the clean end beat the market in fewer than half of one and three year holds. Only over five year stretches did it edge ahead, in about seven runs in ten, and even then by less than the messy end of the same screen delivered all along. The signal was strong. It pointed the wrong way.

That is surprising enough. What comes next is stranger, and it is the real reason this analysis exists.

One backtest is one story, so we built 258 portfolios from the screen, one for every month you could have started, held one, three and five years. Even a screen that lagged had a wide spread: over a single year, eight in ten holds landed between −20% and +75%. Timing swung the outcome as much as the strategy did.

The ranking worked. It just pointed the wrong way.

Most disappointing screens fail because their ranking is noise. This one is the opposite. Sort the market into ten buckets by cash-backed earnings and the returns line up in an almost perfect staircase, a rank correlation around 0.84. The signal is real and strong. The problem is its direction: the bucket with the cleanest earnings made about 8.7% a year, while the bucket with the messiest made closer to 16%. The screen buys the clean end. It bought the losers, reliably.

Why would honest profits lose? Because in India over this decade, the companies converting every rupee of profit into cash were often the mature, finished ones, the cash cows with little left to reinvest. The market paid up instead for businesses ploughing money back in, whose accounts looked messier precisely because they were still growing. Cash-backing did not measure quality here so much as maturity, and maturity was not what paid.

And it charged the highest price of the four

To collect that below-market return you also had to sit through a punishing ride. Take every start date and its worst fall over the following five years: the typical start dropped about 44% from the peak, and the unluckiest near 59%, well past the index's 34%. A screen that underperforms is one thing. A screen that underperforms and hurts more on the way is the combination to be wary of.

What it actually owned

The clean-earnings basket filled up with steady, cash-generative names like Bajaj Auto, Castrol, Pfizer and Symphony, and skewed surprisingly small and mid cap. These are fine businesses. They simply were not where this decade's returns were made, and the screen, by design, walked straight to them.

A right idea can still lose you money

Earnings quality is not a bad concept, and this is not a case against it. It is a case for direction. The same measure that keeps you out of the frauds, when read the other way round, marched this screen into a decade of dull compounders. The lesson is not "ignore cash-backing". It is that a signal can be strong and still be pointed the wrong way, and the only method that tells you which way it points in your market, over your years, is to run it and look. Checking is what turns a respectable idea into a usable one.

So we took it to Krest, and ran it through the whole test.