Narratives and sector rotation
Before you can measure whether a sector is rotating, you have to decide what a sector is. That decision is doing more work than the measurement.
Crypto markets move in themes. Capital and attention concentrate in a category — AI tokens, real-world assets, a particular layer-2 ecosystem — for a period, then move elsewhere. The pattern is real enough that most participants organise their thinking around it.
Measuring it is harder than it looks, and the difficulty starts before any data is collected.
Sector classification is a modelling choice
Equity sector schemes work because there is an underlying fact to appeal to: a company has a primary line of business, and a standards body maintains the taxonomy. Crypto has neither.
Concretely:
- A layer-1 blockchain hosting a large DeFi ecosystem belongs to at least two categories.
- A project that pivots to whatever is currently attracting capital changes category by announcement.
- "AI token" covers projects doing genuine machine-learning infrastructure and projects that added the word to a whitepaper. Both are classified identically.
- Categories are invented as they become popular, so the taxonomy expands in response to the same attention it is meant to measure.
Every sector metric therefore inherits its classifier's judgement. Two providers can show a sector rallying and flat over the same period, both computed correctly, differing only in which assets they placed in it. When a sector chart is shown, the constituent list is part of the result, not a footnote to it.
What sector momentum is built from
Given a classification, sector-level measures aggregate asset-level ones — typically a weighted average of returns, turnover, social volume or sentiment across constituents.
The weighting choice changes the answer substantially:
| Weighting | Describes | Distortion |
|---|---|---|
| Market cap | Where the capital is | One dominant asset becomes the sector |
| Equal weight | Breadth of participation | Micro caps count as much as majors |
| Turnover | Where trading is concentrated | Follows short-term activity, not positioning |
| Social volume | Where attention is | Most exposed to manufactured activity |
A cap-weighted "AI sector" dominated by one large asset mostly reports that asset's behaviour. An equal-weighted version of the same sector can move in the opposite direction on the same day. Neither is wrong; they answer different questions, and the label rarely says which.
Rotation as an observation
Rotation is usually described as capital moving between sectors. What is actually observable is that relative measures changed: one group's turnover and attention rose while another's fell.
That is a weaker statement than the flow language implies. Attention is not conserved — total market attention expands and contracts, so both groups can rise with one rising faster. And measured attention shifting is not the same as capital shifting; those are different observables and they do not always move together.
The defensible formulation is: the distribution of measured activity across categories changed. Whether that reflects positioning or just conversation is a further question, and answering it requires looking at turnover and price alongside the social measures.
The circularity problem
Narrative detection systems identify themes gaining attention, then publish them. Publication attracts attention. Attention is what the system measures.
Any widely consumed narrative tracker partly measures its own output. This does not make such systems useless — being early to a self-reinforcing process still has value — but it means the measurement is not independent of the phenomenon, and a rising narrative score is not evidence that anything outside the attention system changed.
It is the same reflexivity described in social intelligence, operating at category level.
A shape that recurs
Described as a recurring pattern, not a rule with predictive content. The interesting property is that stages two and three look identical while they are happening.
- Origin. A small group discusses a theme. Social volume is low; almost all such themes go nowhere.
- Expansion. Attention broadens, new projects adopt the label, turnover rises across the category.
- Saturation. The theme is widely covered. Projects with tenuous connections attach themselves to it, diluting the category.
- Decay. Attention moves on. The label persists in classification schemes long after the interest does.
Stage three has a measurable signature worth watching: the number of assets classified into a category grows faster than the category's aggregate fundamentals. That is a dilution indicator — and it is a description of what has already happened, not a timing signal.
How Moonboard handles this
The platform groups assets into sectors for its sector map and tracks emerging topics in trending topics, which surfaces themes from news and social sources rather than from a fixed taxonomy.
Two deliberate constraints follow from everything above:
- Sector groupings are presented as a lens, not a fact. They are useful for orientation and should not be treated as authoritative classification.
- No thin page is generated per narrative. It would be straightforward to produce hundreds of category pages for search traffic. They would contain no information, and this domain is built on the opposite proposition.
Open questions
- How much does sector composition choice change measured rotation? Comparing schemes over the same window would quantify it.
- Does the dilution signature — constituent count growing faster than aggregate fundamentals — reliably precede category decay?
- Does narrative attention lead turnover, or follow it?
- Can categories be derived from co-movement rather than assigned by label, and would the resulting groups be more stable?
Related
- Sector rotationThe concept, from first principles
- Social intelligenceThe attention measures underneath
- Ecosystem activityLonger-horizon signals per project
- MethodologyHow asset-level metrics are built