Table 3 Eight-feature architecture for... | Emerald Publishing

Table 3

Eight-feature architecture for crisis detection, organised by feature class with empirically derived crisis thresholds

Feature class	Feature	Crisis threshold
LPPL Physics	Proximity (Hazard Rate)	$> 0.7$ (crash imminence)
	Confidence	$> 0.7$ (11+/15 windows pass)
	Trust	$< 0.3$ (parameter fragility)
Microstructure	Log Return	< −0.04 (heavy-tailed shock)
	Return Acceleration	\|Δr_t\| > 0.05 (inflection)
	Volatility (20-day rolling)	$> 0.40$ annualised
Extremal Events	LMJump Detected	≥2 window sizes flagged
Extremal Events	MD Avg Outlier	MD² > 12.59

Feature class	Feature	Crisis threshold
LPPL Physics	Proximity (Hazard Rate)	$> 0.7$ (crash imminence)
	Confidence	$> 0.7$ (11+/15 windows pass)
	Trust	$< 0.3$ (parameter fragility)
Microstructure	Log Return	< −0.04 (heavy-tailed shock)
	Return Acceleration	\|Δr_t\| > 0.05 (inflection)
	Volatility (20-day rolling)	$> 0.40$ annualised
Extremal Events	LMJump Detected	≥2 window sizes flagged
Extremal Events	MD Avg Outlier	MD² > 12.59