Regime Weather

Crypto market regime: how Regime Weather reads it

Regime Weather treats the crypto market as weather: not as a system to control but as a set of conditions to read. The publication answers, every week:

  • What kind of market is this right now?
  • What changed since the last reading?
  • Where is structural pressure building or releasing?

It does not answer what to buy, when to enter, or what to target.

What “regime” means here

A regime is the dominant set of pressures shaping the market on a given week. It is built from observable, raw features rather than from a model’s classification. Five families of evidence anchor the reading:

  • Leverage and derivatives stress. Funding rates, open interest, basis, liquidations, and crowdedness on perpetual futures.
  • Liquidity and microstructure. Order book depth, spread, imbalance, and fragility across major venues.
  • Cross-venue concordance. Whether spot and derivatives venues agree on direction and magnitude.
  • On-chain and macro context. Settlement activity, holder behavior, and the macro backdrop that confirms or contradicts crypto-native moves.
  • Model and data health. Freshness, coverage, and any degraded inputs used to write the letter.

A regime call is a synthesis of these families, not a single number.

Why raw features and not a model verdict

The Regime Weather letter is built from raw observable features rather than from the regime output of any production trading model. Two reasons:

  1. Defensibility. A letter that cites funding, OI, basis, and depth can be audited against public market data. A letter that cites a model’s regime label depends on that model staying calibrated, which is harder to prove publicly.
  2. Independence from production calibration. When a new model rolls out, its early regime calls are still being validated. The letter should not inherit that uncertainty.

Model health still appears in every letter, but as a status badge: which sources were available, which were degraded, what fallback was used. The regime call itself is not delegated to a model.

What the letter will never publish

  • Entry, stop, or target prices.
  • Specific position recommendations.
  • Proprietary thresholds or model weights.
  • Trade-by-trade execution details.
  • Anything that would let a reader replicate or front-run a position.

These are not editorial choices. They are structural: the letter is built so that nothing inside it can function as a trade signal.

Cadence

One letter per week, on a fixed day. Drafts are produced more often as private calibration material. Real-time alerts are not part of Fase 0; they will only appear once weekly readings have proven that regime changes are detectable and explainable.

How a letter is built

  1. An agent pulls a snapshot of the relevant features from S3.
  2. The agent drafts the letter against a fixed eight-section template.
  3. Anti-advice and redaction policies run as automated checks.
  4. A human reviewer edits and approves.
  5. The letter is rendered to static HTML and published.
  6. A reproducible package of the inputs and derived features is archived.

Every step is auditable. Every step has a fail-closed gate.

How to read it

  • Read the first three sections quickly. They tell you whether the regime changed.
  • Read sections four through six if you want to know which part of the market is doing the work.
  • Read section eight last. It is where uncertainty is named explicitly.

If a letter cannot be written honestly with the data available, it does not ship. Confidence in the publication includes confidence in when not to publish.