Paper: arXiv 2501.08822

Authors: Hamza Bodor, Laurent Carlier

Abstract

The Queue-Reactive model introduced by Huang et al. (2015) has become a standard tool for limit order book modeling, widely adopted by both researchers and practitioners for its simplicity and effectiveness. We present the Multidimensional Deep Queue-Reactive (MDQR) model, which extends this framework in three ways: it relaxes the assumption of queue independence, enriches the state space with market features, and models the distribution of order sizes. Through a neural network architecture, the model learns complex dependencies between different price levels and adapts to varying market conditions, while preserving the interpretable point-process foundation of the original framework. Using data from the Bund futures market, we show that MDQR captures key market properties including the square-root law of market impact, cross-queue correlations, and realistic order size patterns. The model demonstrates particular strength in reproducing both conditional and stationary distributions of order sizes, as well as various stylized facts of market microstructure. The model achieves this while maintaining the computational efficiency needed for practical applications such as strategy development through reinforcement learning or realistic backtesting.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 9.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper presents a sophisticated neural network extension (MDQR) of the Queue-Reactive model, requiring advanced stochastic process theory and deep learning mathematics, while demonstrating strong empirical validation using real market data (Bund futures), reproducing key stylized facts like the square-root law, and emphasizing computational efficiency for practical backtesting.

Research Flowchart

  flowchart TD
  A["Research Goal: Extend Queue-Reactive Model"] --> B["Key Methodology: Multidimensional Deep Queue-Reactive Model"]
  B --> C["Data Input: Bund Futures Market Limit Order Book Data"]
  C --> D["Computational Process: Neural Network Architecture"]
  D --> E["Learn Complex Dependencies &<br>Order Size Distributions"]
  E --> F["Key Findings: Captures Market Impact,<br>Cross-Queue Correlations, &<br>Stylized Facts of Microstructure"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes the Multidimensional Deep Queue-Reactive (MDQR) model, a deep‑learning extension of the Queue‑Reactive framework that relaxes queue independence, adds market features, and models order‑size distributions.

Source sentence
We present the Multidimensional Deep Queue-Reactive (MDQR) model, which extends this framework in three ways: it relaxes the assumption of queue independence, enriches the state space with market features, and models the distribution of order sizes.

Assumes 5

  • Independence between price levels (queue independence) in the original QR model. model
    Source sentence
    Most notably, it assumes independence between price levels and constant order sizes within each level.
  • Constant order sizes within each price level in the original QR model. model
    Source sentence
    Most notably, it assumes independence between price levels and constant order sizes within each level.
  • Only queue sizes are used as features in the baseline QR model. model
    Source sentence
    Both the original QR model and SAQR rely solely on queue sizes as features for order arrival modeling
  • Intra‑day seasonality can be captured by adding the hour of the last event as a categorical feature. market
    Source sentence
    Including hk enables the model to learn intra‑day patterns, such as variations in event intensities across trading hours
Show 1 more
  • Order flow can be classified cleanly into liquidity provision (limit orders) and consumption (cancellations and market orders). data
    Source sentence
    The calibration process requires precise order flow information, which we extract by classifying each update as either liquidity provision (limit orders) or consumption (cancellations and market orders)

Data

  • Euro‑Bund Futures (FGBL) order‑book data order-book unknown
    March to June 2022 · Bund futures (June 2022 contract)
    Source sentence
    The calibration data is derived from Euro‑Bund Futures (FGBL), specifically covering active trading days over three months from March to June 2022 for futures maturing in June 2022

Costs

Not applicable The paper reports no trading or execution result.

Finds 4

  • MDQR reproduces the square‑root law of market impact and cross‑queue correlations observed in Bund futures. statistical
    Source sentence
    Using data from the Bund futures market, we show that MDQR captures key market properties including the square‑root law of market impact, cross‑queue correlations, and realistic order size patterns.
  • Adding hour‑of‑day and previous‑event type features improves log‑likelihood, balanced accuracy, and timing predictions. performance
    Source sentence
    The combination of both hk and ηk−1 leads to the highest log‑likelihood, suggesting these features provide complementary information about order flow dynamics.
  • The MDQR model’s order‑size distribution matches the empirical stationary distribution across the full size range. statistical
    Source sentence
    The left panel displays the stationary distribution of order sizes, where the MDQR‑generated distribution mirrors the historical pattern across the entire range of observed sizes.
  • Incorporating Trade Imbalance features yields a concave market‑impact profile consistent with theoretical expectations, unlike the linear profile of the queue‑size‑only model. performance
    Source sentence
    ... the model incorporating TI features demonstrates a more nuanced and theoretically grounded behavior: it shows a concave‑like progression during the execution phase

Checked by others

Supported by

  • The two square root laws of market impact and the role of sophisticated market participants (2023)
    On the question: Does market impact scale with the square‑root of traded volume/participation rate?
    Compare the findings

    This paper: MDQR reproduces the square‑root law of market impact and cross‑queue correlations observed in Bund futures.

    That paper: During execution, market impact follows a square‑root law in the cumulated traded volume for a given participation rate.

    Both report a square‑root scaling of impact; B reproduces A’s law in Bund futures.

Cited by 3 papers in the archive (3 by independent authors): Neural Estimation of Irreversibility in Real and Simulated … (uses its method or data); Multi-dimensional queue-reactive model and signal-driven … (uses its method or data)

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (33)

  1. A mathematical approach to order book modeling (2013)
  2. Long-time behavior of a Hawkes process–based limit order book (2015)
  3. A Novel Approach to Queue-Reactive Models: The Importance of Order Sizes — → 2405.18594
  4. Statistical properties of stock order books: empirical results and models (2002)
  5. Algorithmic and high-frequency trading (2015)
  6. Conditional Generators for Limit Order Book Environments: Explainability, Challenges, and Robustness — → 2306.12806
  7. A stochastic model for order book dynamics (2010)
  8. Statistical modeling of high-frequency financial data (2011)
and 25 more
  1. Price dynamics in a Markovian limit order market (2014)
  2. A stochastic partial differential equation model for limit order book dynamics (2021)
  3. Limit Order Book Simulation with Generative Adversarial Networks — → 2306.12806
  4. No-dynamic-arbitrage and market impact (2010)
  5. Limit order books (2013)
  6. Simulating and analyzing order book data: The queue-reactive model (2015)
  7. A generative model of a limit order book using recurrent neural networks (2023)
  8. The flash crash: High-frequency trading in an electronic market (2017)
  9. An agent based model of the E-Mini S&P 500 applied to flash crash analysis (2012)
  10. More statistical properties of order books and price impact (2003)
  11. Universal features of price formation in financial markets: perspectives from deep learning (2019)
  12. Statistical theory of the continuous double auction (2003)
  13. Market impacts and the life cycle of investors orders (2015)
  14. Beyond the square root: Evidence for logarithmic dependence of market impact on size and participation rate (2015)
  15. Stylized Facts and Market Microstructure: An In-Depth Exploration of German Bond Futures Market — → 2401.10722
  16. Deeplob: Deep convolutional neural networks for limit order books (2019)
  17. Queue-reactive Hawkes models for the order flow (2019) — link
  18. Does the square-root price impact law belong to the strict universal scalings?: quantitative support by a complete survey of the Tokyo stock exchange market — → 2411.13965
  19. Crossover from linear to square-root market impact (2019)
  20. A theory of power-law distributions in financial market fluctuations (2003)
  21. Get real: Realism metrics for robust limit order book market simulations (2020)
  22. Agent-based modeling in economics and finance: Past, present, and future (2022)
  23. Trades, quotes and prices: financial markets under the microscope (2018)
  24. High-frequency trading strategy based on deep neural networks (2016)
  25. Market microstructure design and flash crashes: A simulation approach (2013)

5 of 33 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.