Rule-Based Market Regime Classification from Intraday Microstructure Measures

Summary

This article combines measurements from an earlier microstructure series into six session labels: Normal, Stressed, Noisy, Informed, Trending, and Mean-Reverting. A priority-ordered rule set applies empirical thresholds to volatility, noise, order flow, clustering, and fractal measures. It also returns a confidence value and a signed directional score, with guidance on adapting position size, stops, or signal filtering to the assigned regime.

The reported thresholds and regime characteristics come from 514 NQ one-minute sessions. The article describes differing sample frequencies and measurement patterns across labels, but does not present the full out-of-sample position-sizing validation; that analysis is attributed to a companion paper. The authors favor transparent rules over a learned classifier and advise recalibrating thresholds for other instruments and periods. Results from one instrument and timeframe may not transfer, and measurement quality limits confidence in the classification.

Key ideas

• The classifier assigns one of six regimes using rules evaluated in a fixed priority order.
• Its inputs combine volatility, noise, order-flow, clustering, and fractal measurements.
• A confidence score reflects measurement reliability, while the composite score follows, fades, or withholds a directional signal by regime.
• Thresholds were calibrated on NQ one-minute sessions and should be recalibrated for other markets and samples.
• The reported study is not itself a complete out-of-sample test of regime-conditioned trading performance.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.