Regime Transition Engine

Features

• Seven-state regime classifier (Noise / Compression / Expansion / Trend Up / Trend Down / Exhaustion / Reversion)
• Minimum dwell + hysteresis debouncing so the state doesn’t flicker on score noise
• Directional Transition Pressure oscillator (−100..+100) reading how close the active state is to flipping
• State-change events plotted at the moment of transition: Compression→Expansion, Expansion→Trend, Trend→Exhaustion, Exhaustion→Reversion
• Optional internal role-score plots (Compression, Expansion, Trend Up/Down, Exhaustion, Reversion) for diagnostics
• Dashboard with current state, next-bias label, transition %, state age, and the underlying volatility/efficiency/variance-ratio readouts

Regime Model

Each bar computes a score (0..1) for all seven candidate states from a shared feature set (volatility percentile, path efficiency, choppiness, linear-regression slope, return z-score, variance ratio, autocorrelation). The state with the highest score becomes the candidate state; the active state only switches to the candidate once it has held for at least the configured dwell period and the candidate score exceeds the active state’s score by more than the hysteresis margin. This keeps the state track stable instead of oscillating bar-to-bar.

Transition Pressure

Once a state is active, a state-specific sub-score estimates how close the market is to moving into its natural successor state (Compression → Expansion, Expansion → Trend, Trend → Exhaustion, Exhaustion → Reversion). This raw pressure is smoothed and signed by trend direction to produce the main oscillator plot — positive values lean bullish/forming, negative values lean bearish/unwinding, depending on the active state.

Dashboard

Top-right table showing: active State (color-coded), Next Bias (plain-language description of the expected transition), Transition % (smoothed pressure), State Age (bars since last switch), and the three core diagnostic readouts (VolPct, Efficiency, Variance Ratio).