Jump Volatility Detector

Quick Reference

Property	Value
Dimension	signal
Category	volatility
Version	v0.9.0 (Beta)
Output Column	jump_flag

Jump volatility detector detects discontinuous price movements

Formula

abs(ret) > (volatility * k)

CDM Inputs

Column	CDM Table	Description
ret	cdm_*	CDM source table
volatility	cdm_*	CDM source table

Parameters

Parameter	Type	Default	Description
k	float [1.0, 10.0]	3.0	Threshold multiplier for jump detection

Output

Column: jump_flag

Binary flag indicating a jump event (1=jump, 0=normal)

Market Intuition & Trading Rationale

Jump detector flags discontinuous price moves: |return| > k × volatility. When a return exceeds k standard deviations of recent volatility, it's classified as a jump — a price move too large to be explained by normal diffusive dynamics. The default k=3 means only returns beyond 3σ are flagged (approximately 0.3% of observations under normality). Real markets produce far more jumps than normality predicts.

Jumps are economically distinct from continuous price moves. They're typically caused by news events, large institutional trades, or liquidity shocks — not by the gradual diffusion of information. A jump signals that something discrete and significant has occurred. The jump_flag output is binary (1=jump, 0=normal), making it a clean event signal for downstream strategies.

The threshold multiplier k controls the sensitivity-vs-specificity tradeoff. Lower k catches more jumps but increases false positives. Higher k only catches extreme events but may miss moderate jumps. The optimal k depends on the instrument's volatility characteristics and the strategy's tolerance for false signals.

Usage Cases

• Jump-robust volatility: Exclude jump observations when computing realized volatility to get a "diffusive-only" vol estimate. Jumps contaminate standard vol estimators with transient spikes. The jump_microstructure_noise feature set uses this approach.
• Event-driven strategies: Trade in the direction of jumps (momentum) or fade them (mean reversion) depending on the instrument's post-jump behavior. Many instruments exhibit jump reversals — the price overshoots and partially reverts within seconds.
• Risk management: During periods of frequent jumps, reduce position sizes and widen stops. Jump clustering (multiple jumps in quick succession) signals extreme market stress — consider pausing trading entirely.

YAML Definition

name: jump_volatility_detector
description: Jump volatility detector detects discontinuous price movements
category: volatility
version: v0.9.0 (Beta)
dimension: signal
status: Pre-release
required_inputs:
  - ret
  - volatility
output_column: jump_flag
output_description: Binary flag indicating a jump event (1=jump, 0=normal)
parameters:
  k:
    type: float
    description: Threshold multiplier for jump detection
    required: false
    default: 3.0
    constraints:
      min: 1.0
      max: 10.0
formula: abs(ret) > (volatility * k)