Radar Perene / Archive / science

Three incidents, one lesson: what fails when a concentrated sector basket breaks

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

Science

Methodologies do not mature in seminars. They mature when something breaks in production — when the series that had been running for months returns a number that makes no sense, and someone has to open the box to understand why. This house's sector-basket methodology carries three such breaks in its history: three real incidents, dated and recorded internally, each followed by a correction that stayed. The details — which companies, which dates, which numbers — belong to the bench. The lesson does not.

In a concentrated market, a sector basket does not break through a formula error: it breaks because the corporate reality of a few companies invades a measure that presented itself as collective. The failure categories are known, repeatable and, once named, verifiable by anyone who keeps such a basket running.

What follows is not the account of the cases — it is the taxonomy they taught.

First category: the member that becomes the group

The most frequent failure is gradual and silent. A basket is born reasonably balanced; over time, one company grows, others shrink or leave, and the weight slides until the "sector average" is, in fact, one company with extras. No alarm goes off, because no rule was violated — the basket keeps calculating exactly what it always calculated. What changed was the meaning of the number, not the number. What detects this is not the formula; it is the periodic audit of the composition, done by someone who has accepted that the question "what does this series measure today?" needs to be asked again from time to time, even when nothing seems wrong.

Second category: the corporate event that swaps the subject

The second failure is abrupt. Mergers, spin-offs, changes of control, sector reclassifications — events that, in a deep market, affect one component among hundreds, and in a shallow market redefine the group overnight. The basket goes on existing under the same name, but the subject of the sentence has changed: the series before the event and the series after it measure different populations. Stitching them together as if they were one produces the kind of invisible discontinuity that contaminates any study run on the series — including retrospective ones, as Track 1 discusses in honest backtesting: a test on a spliced series tests the splice along with it.

Third category: the liquidity that evaporates underneath

The third failure is the most treacherous, because it does not show up in the price — it shows up in the price's absence. Smaller names in a concentrated basket can go days without meaningful trading; the price the series uses is an echo of past sessions. The basket looks calm precisely where nothing is being measured. Under stress, the effect reverses violently: the component that was not trading reprices all at once, and the series records a jump that never corresponded to a real market day. The measurement existed; the thing measured, only intermittently.

What the three have in common

The three categories share an anatomy: none is a calculation error, all are errors of correspondence between what the series claims to measure and what exists to be measured. And all three are invisible to anyone who only looks at the final product. The formula runs, the chart updates, the number comes out — the break lives in the layer the chart does not show.

That is what the three incidents taught the house, and it is why the current methodology treats the maintenance of a basket as continuous work, not initial configuration. The correction applied after each incident became a permanent rule of the process — which rule, exactly, is bench detail; that there are rules born of documented failure is what separates a methodology with scars from a methodology with promises. A sector basket in a concentrated market, the subject of the previous article in this trail, is not an instrument one builds; it is an instrument one watches.

Frequently asked questions

Why not publish the three incidents in detail?

Because the detail would identify specific companies and episodes without adding anything to the transferable lesson. The failure categories are the useful knowledge; the cases are the circumstance in which it was acquired. The full record exists internally, dated.

Do these failures invalidate the house's sector series?

The opposite: each incident produced a correction incorporated into the process, and the post-correction series is more reliable than the earlier one. A methodology with no history of corrected failures either was never tested by reality, or does not record what reality did to it.

Can someone maintaining their own basket guard against these three categories?

The watch questions follow directly from them: is the weight sliding toward a single name? Has a corporate event changed the group's population? Are the smaller components actually trading? None requires sophisticated tooling — they require cadence.

Does this apply to baskets outside Brazil?

It applies to any market where sectors have few liquid representatives. In deep markets the same failures exist in miniature; concentration amplifies them to the point where they stop being noise and become the signal.

---

Continue the trail: Sector rotation: what changes when the sector itself is too small to rotate →

House reading: today's reading is in the Diário; the episodes that stressed whole sectors, in the Atlas.

Reviewing an existing basket in light of these three risk categories is an exercise the house does in conversation, case by case.