Reading a Diagnostics Score Across Regimes

This article is a quantitative analysis piece published for research and educational purposes. It describes how the Diagnostics module's regime layer changes the interpretation of a score; nothing below should be read as guidance to buy, sell, or hold any asset.

Diagnostics ships an optional regime layer (Pro+) that classifies the broader market into risk_on, risk_off, or transition by reading three signals on Bitcoin: annualised volatility, trend R², and the current drawdown state. The classification itself is published in the response so callers can audit the decision. What the layer does — and what makes regime-aware reading load-bearing — is re-weight the factor composites before SQS, FS, and AOS are computed. In risk-off, trend factors are damped and fragility factors are amplified. In risk-on, the reverse. In transition, the layer leans toward symmetric weighting.

Why this matters: the same nominal score means different things in different regimes. Consider an asset that scores 70 AOS in risk-on. Underneath, that 70 is plausibly composed of strong trend quality, decent liquidity, mid-pack volatility, and modestly elevated drawdown. The risk-on weighting flatters the trend signal, so the 70 leans toward 'this asset is participating in the rally and the structural concerns are non-disqualifying.' The same asset scoring 70 AOS in risk-off looks structurally different. The risk-off weighting damps trend and amplifies fragility, so reaching 70 requires materially better fragility numbers — the asset is fragile-resistant, not just trending well. Two equivalent scores; two different stories about what's holding them up.

The corollary that catches readers off guard: a coin's AOS can drift over a window without the underlying metrics changing materially. If the regime detector flips from risk-on to transition, the weights shift, and the same factor composites produce a different AOS. That isn't an artefact — it's the regime-conditional model working as designed. The module reports the regime alongside the score precisely so a reader can avoid mistaking weight-shift drift for genuine deterioration.

Operationally, the recommended reading order is: (1) regime first, (2) AOS as the headline, (3) the SQS / FS pair to see whether the AOS is being driven by quality or by low fragility, (4) the per-factor breakdown if the headline is surprising. The bootstrap confidence intervals (Elite) catch the case where a factor metric is unstable on this asset's data window — wide CIs warn that the headline factor isn't statistically robust on this slice of data.

The module is descriptive throughout. AOS is a measurement of the asset's current structural state under the current regime, not a forecast of price. The regime layer makes the measurement honest about the conditions it's measured under; the boundary between 'measurement' and 'prediction' is the same as it ever was.