Depth Imbalance

Quick Reference

Property	Value
Dimension	execution
Category	order_flow
Version	v0.9.0 (Beta)
Output Column	depth_imbalance

Depth imbalance calculation: (sum(bid_depths) - sum(ask_depths)) / (sum(bid_depths) + sum(ask_depths) + epsilon)

Formula

CDM Inputs

Column	CDM Table	Description
cdm_lob_snapshot.bids	cdm_*	CDM source table
cdm_lob_snapshot.asks	cdm_*	CDM source table

Parameters

Parameter	Type	Default	Description
epsilon	float [0.0, 1.0]	1e-08	Small value to prevent division by zero

Output

Column: depth_imbalance

Depth imbalance ratio between -1 and 1

Market Intuition & Trading Rationale

Depth imbalance measures the asymmetry of resting liquidity: (total_bid - total_ask) / (total_bid + total_ask + ε). Values range from -1 (all depth on ask side — selling pressure) to +1 (all depth on bid side — buying support). Zero means the book is perfectly balanced.

This feature reveals where liquidity providers have committed capital. A positive imbalance means more resting bids than asks — market makers are willing to buy more than sell at current prices, suggesting they expect prices to rise. A negative imbalance means more resting asks — market makers expect prices to fall. Liquidity providers adjust their positioning before aggressive traders arrive, making depth imbalance a predictive signal.

The multi-level computation (extracting bid/ask sizes across all price levels, summing them) captures the full visible depth, not just the touch. A book can appear balanced at the touch but be heavily imbalanced at deeper levels — this feature captures the full picture. The epsilon term prevents division by zero in books where one side has been completely pulled.

Usage Cases

• Pre-trade direction prediction: A positive depth imbalance means more capital committed to buying than selling. Trade in the direction of the imbalance — liquidity providers are positioning for a move in that direction.
• Spoofing detection: Pair with orderbook_update_rate. A rapid change in depth_imbalance combined with high update rates may indicate spoofing (placing and canceling large orders to create false imbalance). Cross-reference with trade_intensity — genuine imbalance should be confirmed by trading activity.
• Execution timing: Execute buy orders when depth_imbalance is positive (more resting bids to absorb your sell). Execute sell orders when depth_imbalance is negative. Your order faces less slippage when there's ample resting liquidity on the opposite side.
• Liquidity stress context: In the depth_imbalance feature set, this feature is paired with cumulative_depth_slope and coefficient_of_variation. Stable imbalance in a deep, consistent book is a high-quality signal. Fluctuating imbalance in a shallow, volatile book is noise.

YAML Definition

name: depth_imbalance
description: 'Depth imbalance calculation: (sum(bid_depths) - sum(ask_depths)) / (sum(bid_depths)
  + sum(ask_depths) + epsilon)'
category: order_flow
version: v0.9.0 (Beta)
dimension: execution
status: Pre-release
required_inputs:
- cdm_lob_snapshot.bids
- cdm_lob_snapshot.asks
output_column: depth_imbalance
output_description: Depth imbalance ratio between -1 and 1
tags:
- depth
- imbalance
- orderbook
- lob
- microstructure
parameters:
  epsilon:
    type: float
    description: Small value to prevent division by zero
    required: false
    default: 1.0e-08
    constraints:
      min: 0.0
      max: 1.0
steps:
- primitive: TRANSFORM
  op: array_extract
  params:
    array: cdm_lob_snapshot.bids
    field: size
    output_col: bid_sizes
- primitive: TRANSFORM
  op: array_extract
  params:
    array: cdm_lob_snapshot.asks
    field: size
    output_col: ask_sizes
- primitive: TRANSFORM
  op: array_sum
  params:
    array: bid_sizes
    output_col: total_bid_depth
- primitive: TRANSFORM
  op: array_sum
  params:
    array: ask_sizes
    output_col: total_ask_depth
- primitive: TRANSFORM
  op: sub
  params:
    output_col: imbalance_numerator
    a: cdm_lob_snapshot.total_bid_depth
    b: cdm_lob_snapshot.total_ask_depth
- primitive: TRANSFORM
  op: add
  params:
    output_col: total_depth
    a: cdm_lob_snapshot.total_bid_depth
    b: cdm_lob_snapshot.total_ask_depth
- primitive: TRANSFORM
  op: add
  params:
    output_col: imbalance_denominator
    a: total_depth
    b: $epsilon
- primitive: TRANSFORM
  op: div
  params:
    output_col: depth_imbalance
    a: imbalance_numerator
    b: imbalance_denominator