Insights

How to Classify Risk-On vs Risk-Off Days

Alphanume Team · April 10, 2026

A daily binary regime signal and how to use it.

A daily risk-on / risk-off classification is the most-basic form of market regime filter. The intent is to produce a single binary label per trading day that summarizes whether the broader US equity market is in a constructive (risk-on) or defensive (risk-off) state. The simplicity is the feature — single-variable classifications are easier to validate and harder to overfit than multi-factor systems.

The simplest defensible approaches

Three classifications that are widely used and defensibly non-overfit:

1. Above/below 200-day moving average. Risk-on when S&P 500 closes above its 200-day moving average; risk-off when below. Trivially simple; reasonably effective.

2. VIX above/below threshold. Risk-on when VIX is below a threshold (e.g., 20); risk-off above. Captures market stress more directly than moving averages but is more responsive to short-term spikes.

3. Combined: above 200-day AND VIX below threshold. Both conditions must hold for risk-on. Reduces false positives but produces more risk-off classifications.

Properties of a good classifier

Whether a classification rule is well-designed depends on:

Stability. Frequent flipping between states suggests the classifier is detecting noise. Good classifications cluster in extended runs of the same state.
Coverage. The two states should both occur with meaningful frequency. A classifier that's risk-on 95% of the time is not providing much information.
Predictive content. Conditional on risk-off classification, subsequent market returns should differ meaningfully from conditional on risk-on.
Robustness. Small changes in threshold should not produce large changes in classification or downstream strategy performance.

Using the classification

Common applications:

Position-size scaling. Full size in risk-on; reduced size (e.g., 50%) in risk-off.
Strategy switching. Different strategies in different regimes.
Risk-overlay hedging. Hedging instruments scaled by regime.
Universe filtering. Some names traded only in certain regimes.

The interpretation discipline

A risk-on / risk-off classification is not a market-direction forecast. The classification describes the current state; it does not predict the next state. Specifically:

Risk-off classification does not mean the market will decline tomorrow.
Risk-on classification does not mean the market will rally tomorrow.
The classification conveys whether current conditions favor or disfavor certain strategy categories.

Common errors

Treating regime classification as a return forecast. The classification is conditional context, not a forecast.
Optimizing thresholds for in-sample performance. Overfit risk.
Using look-ahead in the classification rule. Indicators that need future data are not usable.
Frequent state changes. Trading on each classification flip incurs costs without benefit.

Daily vs longer-frequency classification

Daily classification is the standard but is not always the most useful frequency. Alternative cadences:

Weekly: Smoother, fewer false transitions, less trading.
Monthly: Even smoother, suitable for slower-turnover strategies.
Event-driven: Only re-classify on specific events (volatility breakout, drawdown breach).

For most use cases, daily classification with built-in smoothing (e.g., requiring N consecutive days of breach before flipping) is a reasonable middle ground.

Multi-asset extension

The single-equity-index classification can be extended to multi-asset regime by combining signals across equity, credit, rates, and commodities. Each asset class's regime can be classified individually, with composite regime defined as the joint state.

The trade-off: more inputs increase explanatory power but also overfit risk. For most systematic equity strategies, single-equity-index classification is the better starting point.

For dilution-event short strategies

Risk-off classification tends to coincide with weaker overall equity-market performance, including in the small-cap structural-shorting universe. Strategies that scale back position size or pause activity during risk-on classifications often produce better Sharpe ratios — though sometimes lower absolute returns — than always-on implementations.