RealizedVarianceRegimeDetector

Summary

RealizedVarianceRegimeDetector is RTTA's streaming implementation of: Rolling realized-variance regime detector from squared close-to-close changes.

Update API

result = rtta.RealizedVarianceRegimeDetector().update(close)

The update(...) call consumes one observation using close. advance(...) uses the same inputs when the caller wants to update state without materializing a Python return value.

Theory Of Operation

RealizedVarianceRegimeDetector first constructs a scalar market-state metric from the current observation and compact streaming state, then passes that metric through explicit entry/exit hysteresis. The metric is named in the recurrence below; the hysteresis keeps the output stable until the metric crosses the opposite exit band.

Recurrence

Let \(z_t = close_t\) denote the observation consumed by one update(...) call and let \(\theta\) denote constructor parameters such as window lengths, thresholds, and smoothing constants.

\[\Delta_t=close_t-close_{t-1}, \qquad q_t=\frac{1}{n}\sum_{i\in W_t}\Delta_i^2\]

\[r_t = \begin{cases} 1, & r_{t-1} \le 0 \text{ and } q_t \ge u_e \\ 0, & r_{t-1} = 1 \text{ and } q_t \le u_x \\ -1, & r_{t-1} \ge 0 \text{ and } q_t \le \ell_e \\ 0, & r_{t-1} = -1 \text{ and } q_t \ge \ell_x \\ r_{t-1}, & \text{otherwise} \end{cases}\]

The entry/exit constants satisfy \(\ell_e < \ell_x \le u_x < u_e\).

The return value is the current scalar indicator value.

Implementation Notes

The recurrence is implemented in src/rtta/indicator.cpp in class RealizedVarianceRegimeDetector.

Reference

Background reference