Time Series Momentum Regime

Time series momentum regimes partition historical return sequences into regimes, typically based on volatility clustering, autocorrelation strength, or hidden Markov states, to detect when trend-following or mean-reversion dynamics dominate. In insider-trading surveillance, identifying a persistent-momentum regime elevates confidence in concentrated insider buying signals, since price continuation aligns with documented insider conviction. Conversely, mean-reversion regimes flag potential reversal risk and reduce signal weighting, protecting against false positives from exhausted rallies.

Operationally, regime detection employs rolling windows of 20 to 252 trading days, computing autocorrelation lag-1 or Hurst exponent to classify trending versus mean-reverting states. Insider accumulation during high-momentum regimes scores higher because insider timing synchronizes with market structure; insider transactions during low-volatility mean-reversion windows receive dampened conviction scores, since insider purchases may represent contrarian bets rather than forward-looking conviction. This conditional scoring mechanism reduces whipsaw and improves signal persistence across market cycles.