Signal Half Life
Quick Reference
| Property | Value |
|---|---|
| Dimension | stability |
| Category | statistics |
| Version | v1.0 |
| Output Column | half_life |
Half-life of signal persistence 鈥?time for signal to drop below threshold
Formula
half_life(signal)
CDM Inputs
| Column | CDM Table | Description |
|---|---|---|
signal | cdm_* | CDM source table |
Parameters
| Parameter | Type | Default | Description |
|---|---|---|---|
threshold | float | 0.5 | Threshold for half-life determination |
Output
Column: half_life
Estimated half-life in milliseconds
Market Intuition & Trading Rationale
Signal half-life estimates how long a signal's predictive power persists: half_life(signal). It measures the time (in milliseconds) for the signal's autocorrelation to decay to half its initial value. A long half-life (minutes to hours) means the signal captures persistent information 鈥?once it fires, the edge lasts. A short half-life (milliseconds to seconds) means the signal is ephemeral 鈥?you must act quickly or the opportunity vanishes.
This is the single most actionable stability metric for execution timing. If a signal's half-life is 500ms, you have roughly half a second to execute before the signal's predictive content decays by 50%. If the half-life is 60 seconds, you can afford patient execution. The half-life also tells you the appropriate sampling frequency 鈥?sampling faster than the half-life produces redundant observations; sampling slower misses the signal.
Usage Cases
- Execution urgency calibration: half_life < 1s 鈫?aggressive execution required (market orders). half_life 1s鈥?0s 鈫?moderate urgency (marketable limit orders). half_life > 10s 鈫?patient execution possible (limit orders, schedule over time).
- Signal update frequency: Set your signal update interval to approximately half the half-life. Updating more frequently produces redundant computations; less frequently loses predictive information.
- stability context: Used in
iceberg_detectionpack 鈥?the half-life of the iceberg signal tells you how long the hidden order is likely to persist, informing both entry timing and expected holding period.
YAML Definition
name: signal_half_life
description: Half-life of signal persistence 鈥?time for signal to drop below threshold
category: statistics
dimension: stability
version: v0.9.0 (Beta)
required_inputs:
- signal
output_column: half_life
output_description: Estimated half-life in milliseconds
tags:
- stability
- decay
- statistics
parameters:
threshold:
type: float
description: Threshold for half-life determination
required: false
default: 0.5
formula: half_life(signal)