Insights

What Is a Market Regime Filter?

Alphanume Team · April 11, 2026

Conditioning strategies on risk-on vs risk-off states — what regime filters can and can't do.

A market regime filter is a rule that classifies each historical date into one of several market states — most commonly "risk-on" and "risk-off," sometimes with finer gradations — and conditions strategy behavior on the current classification. The intent is to deploy strategies only in conditions favorable to their underlying logic. The execution requires care: many regime filters look effective in backtests due to overfitting and produce less benefit in live use than expected.

What a regime filter actually does

The mechanic is straightforward:

On each date, compute a regime classification using observable market data.
Adjust strategy behavior based on the classification — e.g., reduce position size in unfavorable regimes, or stop trading entirely.

The intent is to capture the fact that many strategies have regime-dependent performance: short-side strategies often work better in declining markets; volatility-selling strategies often work better in stable markets; momentum strategies often work better in trending markets.

Common regime definitions

Several standard approaches:

1. Volatility-based. Risk-off when VIX or realized volatility exceeds a threshold; risk-on otherwise.

2. Trend-based. Risk-off when the underlying is below a moving average; risk-on when above.

3. Drawdown-based. Risk-off when the market has drawn down more than a threshold from recent highs.

4. Multi-factor. Composite indicators combining volatility, trend, credit spreads, and other macro inputs.

For broad index-level regime classification, the data inputs are public and well-tracked. The choice of threshold is the area where overfitting most commonly enters.

Why regime filters often disappoint

Common reasons regime filters look better in backtest than in live use:

Threshold overfitting. Choosing the volatility or moving-average threshold by optimizing in-sample performance produces optimistic results.
Whipsaw. Regime classifications change frequently in indecisive markets; trading on each classification change generates transaction costs without benefit.
Lag. Most regime indicators are lagging — by the time a "risk-off" classification triggers, the market has already moved. The strategy enters the lower-position-size state after the loss has occurred.
Conditioning on the strategy's own returns. Some regime filters implicitly condition on the strategy's own historical performance, which is circular.

When regime filters help

The applications where regime filters demonstrably add value:

Strategies with strong regime dependence in the structural logic. If the underlying anomaly genuinely only works in certain conditions, filtering on regime improves the realized signal.
Risk management overlays. Reducing leverage or exposure during high-volatility regimes is a risk-management decision that doesn't require timing alpha.
Capital allocation between strategies. Switching between strategy sleeves based on regime can be more effective than filtering any single strategy.

Standard inputs

Common regime-indicator inputs:

VIX level and term structure. Spot VIX, contango/backwardation between 1M and 3M VIX futures.
Realized volatility. Various lookbacks of the underlying index.
Moving averages. 50-day, 200-day on the underlying.
Credit spreads. High-yield CDX, investment-grade CDX.
Yield curve. 2s-10s, 3M-10Y inversion status.
Breadth. Percent of S&P 500 above 200-day moving average.

Design discipline

To reduce overfit risk in regime filter design:

Specify the filter before testing. Choose thresholds based on logic, not on backtest performance.
Use simple filters. Multi-factor regime indicators have more degrees of freedom and more overfit potential.
Cross-validate. Test on hold-out periods.
Compare to no-filter baseline. The benefit of the filter should exceed the cost of the increased complexity.
Watch for whipsaw. Count the number of regime flips in the backtest period. High flip rates suggest the filter is detecting noise.

For dilution-event short strategies

Dilution-event short strategies tend to perform better in risk-off regimes (when broader market is weak) and worse in strong risk-on regimes (when even structurally weak issuers can rally on broad market strength). A regime filter that scales back position size during low-VIX, strongly-rising markets and restores it in higher-VIX or declining markets is a defensible application.