Comparing Probabilistic Queue Models in Market-Making Backtests

Summary

This tutorial examines how probabilistic queue-position assumptions affect simulated limit-order fills and market-making results. It implements a grid quoting strategy based on a GLFT-style market-making model, estimates order-arrival intensity from observed trade depths, fits a log-linear relationship, and updates volatility estimates over rolling data windows. Inventory influences the reservation price, while estimated spread and volatility determine quote spacing. The backtest is run under three queue models for a basket of futures assets.

The comparison uses recorded equity and risk-adjusted statistics to visualize outcomes across the models. The example therefore illustrates a useful calibration workflow: compare fill assumptions with live trading behavior before trusting simulated performance. Its evidence is limited to the provided historical sample and setup; the excerpt does not report numerical outcomes in prose or establish that any model is generally superior. The assumed order-arrival distribution may not fit every asset, and the analysis filters assets by a performance-based criterion. Results also depend on the stated fee and rebate assumptions, latency data, fill settings, and strategy parameters.

Key ideas

• Queue-position assumptions change simulated fill rates and can materially affect market-making backtests.
• The example estimates trade-arrival intensity from observed market-order depths and fits a log-linear model.
• A grid quoting strategy adjusts its reservation price for inventory and uses volatility in setting spreads.
• Three probabilistic queue models are compared across multiple futures assets using equity and risk-adjusted statistics.
• The conclusions are limited by asset selection, distribution assumptions, fees, latency, and the historical test setup.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.