Microstructure Noise Ratio

Quick Reference

Property	Value
Dimension	quality
Category	volatility
Version	v1.0
Output Column	noise_ratio

Microstructure noise ratio: realized_vol(short_window) / realized_vol(long_window) 鈥?estimates noise contamination

Formula

rolling_std(ret, short_window) / rolling_std(ret, long_window)

CDM Inputs

Column	CDM Table	Description
ret	cdm_trade_enriched	Trade data enriched with bar context 鈥?price, volume, side, trade type

Parameters

Parameter	Type	Default	Description
short_window	integer	1000	Short window for volatility estimation
long_window	integer	60000	Long window for volatility estimation

Output

Column: noise_ratio

Ratio of short-term to long-term realized volatility

Market Intuition & Trading Rationale

Microstructure noise ratio measures how much of short-term price variance is noise vs signal: realized_vol(short) / realized_vol(long). At very short horizons (milliseconds to seconds), price changes are dominated by microstructure effects 鈥?bid-ask bounce, order book dynamics, tick size discretization 鈥?rather than fundamental information. At longer horizons (minutes+), information dominates.

A noise ratio > 3 means short-term variance is triple the long-term variance 鈥?the market is extremely noisy at high frequency, and HFT strategies will struggle to extract signal from the noise. A noise ratio between 1鈥? means short and long-term volatility are comparable 鈥?the market is relatively clean, and short-horizon signals are more reliable. A noise ratio < 1 (contango in volatility term structure) means long-term uncertainty exceeds short-term 鈥?typical during calm periods with looming event risk.

Usage Cases

• HFT viability assessment: noise_ratio > 3 鈫?avoid HFT strategies; the signal-to-noise ratio at high frequency is too poor. noise_ratio < 2 鈫?HFT strategies have a reasonable chance of extracting signal from noise.
• Volatility term structure: The ratio is effectively the slope of the volatility term structure at short horizons. Pair with volatility_term_structure_slope (stability dimension) for a complete vol term structure profile.
• quality context: Used in jump_microstructure_noise and realized_volatility_spectrum packs 鈥?high noise_ratio means short-term vol is dominated by noise, so short-horizon realized vol estimates are unreliable.

YAML Definition

name: microstructure_noise_ratio
description: 'Microstructure noise ratio: realized_vol(short_window) / realized_vol(long_window)
  鈥?estimates noise contamination'
category: volatility
dimension: quality
version: v0.9.0 (Beta)
required_inputs:
  - ret
output_column: noise_ratio
output_description: Ratio of short-term to long-term realized volatility
tags:
  - quality
  - noise
  - volatility
parameters:
  short_window:
    type: integer
    description: Short window for volatility estimation
    required: false
    default: 1000
  long_window:
    type: integer
    description: Long window for volatility estimation
    required: false
    default: 60000
formula: rolling_std(ret, short_window) / rolling_std(ret, long_window)