Regime Detection: Classification, Not Prediction
The Distinction That Matters
OVRWCH does not predict where markets are going. It classifies where markets are.
A predictive system claims future knowledge: “the S&P 500 will decline 10% in the next 60 days.” That claim requires the model to know something about future returns — a demonstrably difficult and often fraudulent enterprise.
A classification system claims present knowledge: “current market conditions match the statistical profile of environments that have historically preceded extended drawdowns.” The claim is about now, not tomorrow. The portfolio responds to the classification, but the model is not asserting directional edge.
This distinction changes what the system can be wrong about. It can misclassify the current state. It cannot be wrong about a prediction it never made.
Four States
OVRWCH’s regime detector identifies four distinct market environments from a proprietary set of macro features. The states are learned from data, not hand-tuned with rules or thresholds.
Over 18 years of daily data (2008–2026):
| State | Frequency | Character |
|---|---|---|
| Expansion | ~51% of days | Low volatility, broad participation, positive momentum |
| Contraction | ~38% of days | Mixed signals, narrowing breadth, directional uncertainty |
| Crisis | ~11% of days | Elevated volatility, negative momentum, credit stress |
| Transition | <1% of days | Rapid regime shifts, low classification confidence |
Each state triggers different allocation behavior. The portfolio is not making a directional bet — it is matching its composition to the observed environment.
What the System Detected in April 2026
On April 10, 2026, OVRWCH classified the current environment as Crisis with high confidence. The portfolio automatically positioned defensively: heavy allocation to safe-haven assets (sovereign bonds, gold, yen) with minimal equity exposure.
Whether the classification proves correct in hindsight is secondary. The system’s job is to match the portfolio to the environment it observes. If the environment shifts, the portfolio shifts with it.
The False Positive Problem
Not every Crisis classification precedes a real drawdown. Across 18 years, the detector has identified Crisis-level conditions multiple times that resolved without significant market decline. This is a known and documented limitation.
The cost of false positives is missed upside — the portfolio deleverers defensively while the market recovers. OVRWCH addresses this through a confirmation mechanism that requires elevated conditions to persist before full defensive positioning activates. The specific parameters of this mechanism are proprietary.
The tradeoff is explicit: accept occasional missed upside in exchange for dramatically reduced exposure during real drawdowns. Over full market cycles, this tradeoff has historically favored the defensive posture.
This note describes OVRWCH’s general approach to regime detection. Specific features, model parameters, and implementation details are proprietary.
Get OVRWCH's regime report and trade analysis.
Free. No spam. Unsubscribe anytime.
We'll connect this to Beehiiv when we launch.