How a book becomes
a backtest.
Investment writing is persuasive by design. A backtest is not. Taking a rule out of a book and putting it into a program strips away the rhetoric and leaves the claim exposed.
The four steps
Identical for every issue, so results stay comparable.
Pick the rule
Not the book — the rule. One sentence the author actually wrote, specific enough that a machine could follow it and a reader could disagree with it.
Write it down precisely
Universe, entry condition, position sizing, exit. Every ambiguous word in the original sentence has to be resolved into one concrete choice, and the choice gets published.
Run it over a decade
The program runs on historical market data, quarterly or monthly depending on the rule. Costs and turnover are recorded. Nothing is tuned after the fact.
Publish what came back
Results are published as they are. A rule that failed is a valid issue. A rule that only worked because of one setting is more interesting than one that worked everywhere.
House rules
Four commitments that decide what gets published and how it is written.
One rule per issue
Whole books are too vague to test. Isolating a single rule keeps the claim falsifiable.
The original definition, first
The author gets to state the rule in their own terms before we translate it. Translation is where most distortions happen.
No hindsight tuning
Parameters are fixed before the run. If a version was tested and discarded, that belongs in the write-up, not in the bin.
Failures are published
A library of only-working ideas is a marketing brochure. The crate keeps the duds.
What a backtest can and cannot tell you
A backtest shows that a rule, applied mechanically to a specific universe over a specific window, produced a specific distribution of outcomes. It does not show that the rule captures a durable economic truth, and it never shows what will happen next.
Some of the effects we find are artifacts: survivorship in the data, look-ahead in a reported figure, the ex-date price gap that never recovers. We write these down in the issue rather than cleaning them up, because a reader who cannot see the seams cannot judge the result.
Why we publish with the losses in
The most useful thing a library of this kind can do is show you which famous ideas did not survive contact with data. That information is only available if failures are as publishable as wins.
What this is not
This is educational research. Nothing here is investment advice, a recommendation, or an offer to buy or sell any security. A rule that worked in a backtest is not a strategy we are telling you to run.