The map · Prologue

Twenty-six tests, and what was left

The claimA method built from good books was tested on history, most of it did not hold, and the rest of this book is what was left.

About 4 minutes · the investor's route · the builder's route.

This book describes a slow way of buying shares, and the small program that keeps the person doing it honest. It was written twice. The first time it described a method. A few weeks later it had to describe what was left of one.

The method took three months to build. It came from the books anyone would start with: Benjamin Graham, Warren Buffett's letters, Joel Greenblatt, the academic papers on value and quality. Every rule had a reason and most had a citation. By late October 2026 there were about 150 written rulings. A company had to have fallen 15 to 50 percent from its high. The survivors were ranked so the best came first. Seven warning signs ended a reading on the spot. Every holding had a price at which it would be sold without argument. Every value had a cushion under it. Four quality tests were borrowed from the way Buffett describes a good business. It read well. It felt careful.

Reading well is not evidence. So between 29 October and 9 November the whole method was put on trial against history: the United States from 2016 to 2020 and the Nordic markets from 2010 to 2021. The companies that later went bankrupt or were bought were put back in, and every figure was allowed in only on the day it became public. Each test was written down before it ran: what is claimed, what is measured, where the bar is, and what happens to the rule if it fails. There were twenty-six attempts.

What happened

Most of the rules died.

• The price stop. No level helped, and with trading costs every one of them cost money.
• The 15 to 50 percent band. The companies inside it did no better than the ones outside.
• The ranking. It did not put the better companies first. The list is now in a fixed random order, and says so.
• Three of the seven warning signs. Falling sales, a squeezed margin and swelling unpaid invoices did not separate the companies that did well from those that did not.
• The debt limits, the quality tests as a reason to drop a company, and the cushion as a binding buy line. None separated returns. The tests and the cushion are still printed, as information.

What survived is short: the valuation itself, the rule that figures come only from the company's own reports, the belief about growth written down before the price is looked at, positive spare cash, at most six percent of the money in one company, and the reading a person does by hand.

And the trial did not find what every investor hopes a trial will find. No edge has been shown. In the Nordic markets the cheaper companies by the method's own value beat the dearer ones in most of the years it was developed on. A check on two later years, written down in advance and run once, did not confirm it. Against four well-known published value methods, this one won none of twelve comparisons.

Why go on, then

Three reasons, none of them hopeful in the usual way.

The money is somewhere else. A broad index fund is the base of the owner's savings. This method is the smaller part: the part that is studied, not trusted. Nothing in this book asks you to trust it with more.

What is left is the part no history can test. Reading a company's own reports. Saying in one sentence what it sells and why its customers come back. Writing down what you believe before the market tells you what to believe. These are tested the slow way: every company the method says no to is written down with its price that day and measured against the index for years (chapter 12). The verdict on the hand-made parts arrives around 2030.

The taking apart is the useful thing. A person with a method and no test has a belief. This is a record of what it costs to find out: the rule that every test is written down first, the faults found in our own data, and the day a value fell by a third because a rule was finally applied to a company that had already been bought.

How to read it

Part One is the method as it stands now. Each chapter states its claim in one sentence, tells the rule, and says what the test found where there was one. The tests have their own chapter in the middle. Part Two is the machine, and the faults it has caught in itself.

One real company runs through the whole book in a box marked CENTA. The owner bought it, valued it wrongly, valued it again and sold it, all within eight days in October 2026. It shows each rule doing its work, or arriving too late.

less reference/backtest/WHAT-WE-TESTED.md

The slow way, made survivable. Built 2026-11-09 from the repository as it stood that day. NAME A and NAME B are fictional. One real company, CENTA, is followed through the book, with figures from its own filings and the owner's own trades. Not financial advice: one private investor's method, shared as it is.