MarketGraphsThe Lens · MethodologyAnomaly (hybrid)
← all engines

Anomaly (hybrid)

What's not behaving like its peers OR moving the cross-asset state?
ExperimentalVerdict · EXPERIMENTAL
Alternative view · method variant · anchor metric: peer-pct ⊕ drift-detector
What it measures

A hybrid of the peer-percentile axis and a Mahalanobis cross-asset state-distance detector — flags an instrument if it diverges from its cohort or if it's moving the cross-asset state. Wider net than peer-pct alone.

\text{score} = \max(\text{peer\_pct\_divergence},\ \text{mahal\_state\_dist})
\text{union of the two candidate sets, de-duplicated}
\text{forward\_window\_test} = 5 \text{ sessions}

Catches two flavours of abnormal: the lone outlier, and the instrument quietly dragging the whole cross-asset picture. Higher recall on the union, but noisier — runs as a comparison method, not the headline.

Limitations
  • Method variant — graded alongside peer-pct, not independently shipped as a headline read.
  • Union noise — the max-of-two logic admits more false positives than the percentile axis alone.
In the product

Runs beside anomaly_peer_pct in the Anomalies chapter — both flag instruments diverging from their cohort; the page shows peer-pct as primary.

Grading & stability
Gate rate
loading…
Stability / churn
not emitted
Mean confidence
loading…
fetching /api/panel/about/lens_engine.anomaly_hybrid/W/all …

Graded on the Tier-1 regime-test harness — curated macro events × 18-yr backfill × recall@K. Verdict EXPERIMENTAL (method variant).

Charts it reads
Relative rotation graph
rrg_scatter
Major assets
major_assets
Anomaly (hybrid) · The Lens methodology← all engines