Open the Magic Formula's engine and there is almost nothing inside. The ranking barely ranked, and trusting it hardest cost the most.

Every ranking strategy makes one implicit promise: that its order means something, that its first choice is better than its fiftieth. Test that promise directly and the trouble shows immediately.

The celebrated ranking is close to random

Split the market into ten buckets by the formula's own combined score and look at what a decade paid each one. The bucket the formula ranked worst earned 17.5% a year. Its favourite earned 16.4%. The best performer of all was the eighth bucket, one the formula considered mediocre. The statistical relationship across all ten is close to zero. The two ingredients are sensible; their sum, in this market, sorted almost nothing. The labels on the ten buckets changed. What was inside them barely did.

The faithfulness penalty

Here is where a random ranking turns costly. If the order means nothing, then concentrating into its top picks buys nothing but concentration risk, and the data delivers exactly that verdict, in grades. The broad top decile earned 16.4% a year. The top thirty, 14.8%. And the ten stocks the formula loved most earned 8.3% a year through a 75% fall, beating the index in just 2 of every 100 five year windows. Follow the formula loosely and you roughly matched the market. Follow it to the letter and it halved your return and deepened your worst fall. Faith in the ranking was the single most expensive input an investor could add.

Positive is not the same as better

One number in the formula's favour deserves honest airtime, because it explains why believers stay believers. Held five years, 99 of 100 starts made money. Almost nobody who ran this formula patiently ended up poorer, and that feels like vindication from the inside. But across those same windows, fewer than half beat the index fund that required no formula, no rebuilding and no ranking. The formula made its followers money in a market that was making everyone money, and the difference between those two sentences is the entire case for testing.

Every cut of the formula shows the same signature: near certain to be positive, coin toss to be better. A strategy can feel like it is working every single year and still be adding nothing at all.

The weather, briefly

The formula's months follow the familiar small cap seasons, better when the economy expanded, better when rates eased, near flat when growth stalled. Nothing here is unusual, and nothing here rescues it: the winners we have tested shared the same weather and cleared the index through it. The problem was never the climate. It was the engine.

What it really is, underneath the name

Ask what the Magic Formula portfolio most resembles in the real world and the answer is plain: a broad basket of small Indian companies, tracking the equal weighted market closely, with a hair less return than simply owning that market would have paid. Whatever the two famous ranks contributed, the humble act of holding lots of small companies contributed more, and the ranking's net addition on this decade rounds to nothing. The name promises a formula that beats the market. The data shows a formula that rode the market, under a better title.

What a reader should take from this

Three things, none of them a variant to buy. First, a ranking must be tested against its own deciles before it deserves conviction; this one fails that test, and the failure was invisible from the outside. Second, whenever you do run a screen whose sort you have not verified, breadth is self defence, the broad decile beat every tighter cut here precisely because it diluted the ranking. Third, and largest: the single most famous number in popular investing concealed a coin toss, a 68% fall and a ranking that barely worked. If that number needed this much checking, so does every other number you have ever been sold. That is not cynicism. That is the whole reason testing exists.

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