RTTADocumentation Built by Goblin Reactor ↗

CalibrationDriftDetector

Incremental, causal technical analysis documentation

Summary

CalibrationDriftDetector is RTTA's streaming implementation of: EWMA probability-calibration error drift detector.

Update API

result = rtta.CalibrationDriftDetector().update(probability, outcome)

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

Theory Of Operation

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

\[e_t=|\mathbf{1}[outcome_t>0]-\operatorname{clip}(probability_t,0,1)|\]
\[q_t=\alpha e_t+(1-\alpha)q_{t-1}\]
\[r_t = \begin{cases} 1, & r_{t-1} = 0 \text{ and } q_t \ge e \\ 0, & r_{t-1} = 1 \text{ and } q_t \le x \\ r_{t-1}, & \text{otherwise} \end{cases}, \qquad x < e\]

The return value is the current scalar indicator value.

Implementation Notes

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

Reference