Radar Perene / Archive / science

Market relationships that do not survive: what remained after testing everything again

◦ Index methodology v2.2 (working papers with DOI). See the methodology.

Science

Every market accumulates a folklore of gears: when A rises, B falls; sector X leads the index; the ratio between Y and Z "always" reverts. These relationships circulate on trading desks and in reports as if they were physical properties of the system — inherited, repeated, rarely dated and almost never retested. The Brazilian market has its own inventory of them. This house decided to treat that inventory the way a scientist treats a collection of hypotheses: putting all of it, at once, before the same ruler.

The working paper Brazilian Intramarket Relationships started from a set of relationships between sectors and asset classes of the Brazilian market and submitted them to systematic retesting. The outcome, published in version 2.0 with a permanent DOI: nearly all of them failed to hold — and the single surviving relationship was the one linking the utilities sector to the market's defensive behavior. The current version openly revises a finding the previous version upheld, with both preserved in the repository. The paper is, at once, a result about the market and a public record of the house's audit protocol in operation.

Why folklore manufactures so many false gears

The high mortality should surprise no one who knows the statistics of the problem. Financial series are the perfect habitat for the illusory relationship: whoever crosses dozens of sectors and asset classes against each other has, by arithmetic alone, hundreds of candidate pairs — and chance guarantees that a fraction of them will display convincing synchronies in any finite sample. Track 1 describes the mechanism in spurious regression: persistent series, when compared, manufacture statistical association where no link exists. Market folklore is, in large part, the social residue of that mechanism — the coincidences someone noticed, named and put into circulation, where they survive by repetition rather than by evidence.

Add to that the Brazilian sample problem, documented in the previous trail: few assets, short history, concentrated sectors. The environment that most manufactures illusory relationships is exactly the one that offers the least material to refute them.

What the retest did — and what the paper tells

The design of the retest battery — which variations, in what order, with what cut-off criteria — is bench work, and the paper doses what it exposes. What it publishes is what matters to the reader: the starting inventory, the common ruler and the outcome. And the outcome has two levels. The first is the empirical result: of the relationships that folklore and the study's first version treated as candidates, one survived. The second level is the methodological one, and it is what earns the paper its place in this trail: version 2.0 records, in its own text, the revision of a finding version 1.1 upheld. The document does not inherit the earlier conclusion in silence — it unsays what needed unsaying, dated, with both versions in view.

It is the same cycle this trail documented for the ILI: detect, correct, record. There, the re-examination reached an index in production; here, it reached the inventory of relationships that informs the house's intermarket reading. The ruler spares no shelf of the collection — and the public record of each application is what turns the previous sentence from advertising into verifiable fact.

The value of the cemetery

The editorial temptation, in a study like this, would be to publish only the survivor — "the relationship that works in the Brazilian market" would make a more comfortable title. The paper did the opposite, and the choice carries the study's most useful information. For anyone who uses intermarket relationships, the list of what fell is worth more than the list of what remained: each refuted relationship is one gear of the folklore that stops operating silently inside other people's analyses. And the size of the mortality calibrates the credit owed to the survivor — a relationship that crossed the ruler that felled the others carries a standing no untested relationship can claim.

There is also the rarer, more valuable second-order effect: a study that publishes its own revision teaches the reader how to treat the house's entire production. No finding here is definitive by decree; it is in force until the next retest. What remained after testing everything again was not just one relationship — it was the demonstration, with a DOI, that "testing everything again" is a sentence with operational content in this house.

And the survivor?

The relationship between utilities and the defensive behavior of the Brazilian market deserves more than the closing paragraph of an article about mortality — what makes it more stable than the others, and what that stability says about the structure of the local market, is the subject of the trail's next article.

Frequently asked questions

Which relationships were discarded?

The public paper doses that inventory at the measure that serves it, and this article does not go beyond it. The transferable content is in the published outcome: near-total mortality, one survivor, and the revision recorded between versions — all verifiable in the repository.

Can a relationship refuted in the study become valid again?

It can — refutation under one ruler is not eternal banishment, and regimes change. What the study establishes is that, under systematic retesting in the sample examined, the relationship did not hold. Whoever wants to rehabilitate it inherits the burden of proof, now with a public reference to argue against.

Doesn't publishing the revision of one's own finding erode the study's credibility?

It erodes the credibility of those who treat findings as dogma. Under a frozen-version regime, the declared revision is the rational behavior and the cheapest to audit: both versions sit in the repository, and any reader can check what changed. The alternative — the fragile finding kept in silence — is the one that charges interest.

Does the result apply to other markets?

The empirical outcome belongs to the Brazilian market, in the sample examined. What travels is the method and the base lesson: inventories of inherited relationships, in any market, tend to shrink drastically under systematic retesting — and inventories that never shrink are inventories that were never retested.

---

Continue the trail: Utilities and the rest of the market: the relationship that passed the robustness test →

House reading: today's intermarket reading is in the Diário; the episodes where the folklore's gears failed in public, in the Atlas.

Retesting a specific market relationship under the same ruler — before it decides an entire analysis — is work the house carries out on request, in conversation.