Inferensys

Glossary

Slippage Model

A simulation component that calculates the difference between the expected price of a trade and the price at which the order is actually filled due to latency and market movement.
Performance engineer optimizing AI latency on laptop, latency charts visible, technical optimization session.
EXECUTION SIMULATION

What is a Slippage Model?

A slippage model is a quantitative simulation component that calculates the divergence between an order's intended execution price and its actual fill price, accounting for latency, liquidity constraints, and adverse market movement.

A slippage model is a simulation component that calculates the difference between the expected price of a trade and the price at which the order is actually filled. It quantifies execution degradation caused by market impact, latency, and bid-ask spread dynamics during the interval between signal generation and order completion within a backtesting engine.

These models typically ingest order book depth, tick-level volume profiles, and volatility regimes to estimate fill probability at each price level. By incorporating a slippage model, quantitative developers prevent look-ahead bias and generate realistic implementation shortfall estimates, ensuring backtested strategies reflect genuine execution constraints rather than idealized theoretical prices.

TRANSACTION COST ANALYSIS

Core Components of a Slippage Model

A slippage model quantifies the divergence between a theoretical decision price and the actual execution price. The following components decompose this cost into its constituent parts for accurate simulation.

01

Bid-Ask Spread Cost

The immediate cost of crossing the spread to execute a market order. This component captures the difference between the mid-price and the prevailing bid or ask.

  • Quoted Spread: The difference between the best bid and ask at order arrival.
  • Effective Spread: Double the difference between the trade price and the mid-point at the time of order entry.
  • Realized Spread: Measures the market maker's revenue after accounting for adverse price movements post-trade.

In a backtesting engine, this is often modeled as a fixed half-spread penalty applied to every aggressive order.

0.01%–0.50%
Typical Spread Cost Range
02

Temporary Market Impact

The transient price concession required to attract liquidity and fill an order. This cost decays rapidly after execution completes as the order book replenishes.

  • Linear Impact Models: Assume impact scales proportionally with order size relative to average volume.
  • Square-Root Models: A non-linear relationship where impact scales with the square root of participation rate, widely validated in empirical microstructure research.
  • Resilience Decay: The speed at which the price reverts to its undisturbed level, typically modeled as an exponential decay function.

This component is critical for sizing child orders in execution algorithms like TWAP and VWAP.

√(Q/V)
Square-Root Scaling Law
03

Permanent Market Impact

The irreversible price movement caused by the information content of a trade. The market interprets aggressive buying or selling as a signal of informed trading and adjusts the equilibrium price accordingly.

  • Information Leakage: The degree to which an order reveals private alpha, permanently shifting the fair value.
  • Kyle's Lambda: A canonical measure of permanent impact, defined as the price change per unit of net order flow.
  • Cumulative Effect: Unlike temporary impact, this component does not decay and accumulates over the life of a meta-order.

Accurate permanent impact modeling is essential for strategies with large directional positions or long execution horizons.

5%–35%
Permanent Share of Total Impact
04

Delay Cost

The adverse price movement that occurs between the decision time and the initial order arrival due to latency, processing, or intentional scheduling delays.

  • Latency Slippage: Price drift during network and system processing, critical for high-frequency strategies.
  • Scheduling Drift: The cost of waiting to execute a slice in a scheduled algorithm, often modeled as a Brownian motion with drift.
  • Opportunity Cost: The risk that an unfilled limit order misses a favorable price move, a direct trade-off against spread capture.

Delay cost is the primary driver of the implementation shortfall metric used to evaluate execution quality.

σ√(Δt)
Delay Volatility Scaling
05

Order Book Resilience

The rate at which depleted liquidity replenishes after a trade. A resilient book recovers quickly, minimizing the cumulative impact of sequential orders.

  • Resilience Parameter: The exponential decay constant governing how fast the order book returns to its steady-state shape.
  • Hawkes Process Models: Self-exciting point processes that capture the clustering of order book events and liquidity regeneration.
  • Queue Position Effects: The interaction between order cancellation rates and new limit order arrivals at each price level.

Backtesting engines with order book replay capability can directly measure resilience rather than relying on parametric approximations.

< 1 sec
Typical Resilience Half-Life
06

Adverse Selection

The cost incurred when trading against counterparties with superior information. This component captures the probability that a fill occurs precisely when the market is about to move against the position.

  • Pre-Trade Toxicity Metrics: VPIN (Volume-Synchronized Probability of Informed Trading) estimates the likelihood of informed flow.
  • Post-Trade Drift: The average price movement following a fill, used to calibrate adverse selection in limit order models.
  • Spread Capture Trade-Off: Passive orders earn the spread but suffer adverse selection; the model must balance these competing forces.

Adverse selection is the primary reason limit orders are not risk-free and must be modeled stochastically.

SLIPPAGE MODELING

Frequently Asked Questions

Clear answers to common questions about simulating execution shortfall in backtesting engines.

A slippage model is a simulation component that calculates the difference between the expected price of a trade and the price at which the order is actually filled due to latency and market movement. It works by applying a mathematical function to the theoretical execution price, typically incorporating factors such as order size relative to available liquidity, the current bid-ask spread, and volatility. In a backtesting engine, the model intercepts the fill logic and adjusts the fill price downward for buy orders or upward for sell orders, degrading the simulated performance to reflect real-world execution costs. Advanced models decompose slippage into temporary market impact (which reverts) and permanent impact (which shifts the equilibrium price).

Prasad Kumkar

About the author

Prasad Kumkar

CEO & MD, Inference Systems

Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.

His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.